| P2PSIP | C. Jennings |
| Internet-Draft | Cisco |
| Intended status: Standards Track | B. B. Lowekamp, Ed. |
| Expires: September 16, 2011 | Skype |
| E.K. Rescorla | |
| RTFM, Inc. | |
| S.A. Baset | |
| H.G. Schulzrinne | |
| Columbia University | |
| March 15, 2011 |
REsource LOcation And Discovery (RELOAD) Base Protocol
draft-ietf-p2psip-base-13
This specification defines REsource LOcation And Discovery (RELOAD), a peer-to-peer (P2P) signaling protocol for use on the Internet. A P2P signaling protocol provides its clients with an abstract storage and messaging service between a set of cooperating peers that form the overlay network. RELOAD is designed to support a P2P Session Initiation Protocol (P2PSIP) network, but can be utilized by other applications with similar requirements by defining new usages that specify the kinds of data that must be stored for a particular application. RELOAD defines a security model based on a certificate enrollment service that provides unique identities. NAT traversal is a fundamental service of the protocol. RELOAD also allows access from "client" nodes that do not need to route traffic or store data for others.
THIS DOCUMENT AND THE INFORMATION CONTAINED THEREIN ARE PROVIDED ON AN “AS IS” BASIS AND THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST, AND THE INTERNET ENGINEERING TASK FORCE, DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION THEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 16, 2011.
Copyright (c) 2011 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
This document may contain material from IETF Documents or IETF Contributions published or made publicly available before November 10, 2008. The person(s) controlling the copyright in some of this material may not have granted the IETF Trust the right to allow modifications of such material outside the IETF Standards Process. Without obtaining an adequate license from the person(s) controlling the copyright in such materials, this document may not be modified outside the IETF Standards Process, and derivative works of it may not be created outside the IETF Standards Process, except to format it for publication as an RFC or to translate it into languages other than English.
This document defines REsource LOcation And Discovery (RELOAD), a peer-to-peer (P2P) signaling protocol for use on the Internet. It provides a generic, self-organizing overlay network service, allowing nodes to efficiently route messages to other nodes and to efficiently store and retrieve data in the overlay. RELOAD provides several features that are critical for a successful P2P protocol for the Internet:
These properties were designed specifically to meet the requirements for a P2P protocol to support SIP. This document defines the base protocol for the distributed storage and location service, as well as critical usages for NAT traversal and security. The SIP Usage itself is described separately in [I-D.ietf-p2psip-sip]. RELOAD is not limited to usage by SIP and could serve as a tool for supporting other P2P applications with similar needs. RELOAD is also based on the concepts introduced in [I-D.ietf-p2psip-concepts].
In this section, we provide a brief overview of the operational setting for RELOAD. See the concepts document[I-D.ietf-p2psip-concepts] for more details. A RELOAD Overlay Instance consists of a set of nodes arranged in a connected graph. Each node in the overlay is assigned a numeric Node-ID which, together with the specific overlay algorithm in use, determines its position in the graph and the set of nodes it connects to. The figure below shows a trivial example which isn't drawn from any particular overlay algorithm, but was chosen for convenience of representation.
+--------+ +--------+ +--------+
| Node 10|--------------| Node 20|--------------| Node 30|
+--------+ +--------+ +--------+
| | |
| | |
+--------+ +--------+ +--------+
| Node 40|--------------| Node 50|--------------| Node 60|
+--------+ +--------+ +--------+
| | |
| | |
+--------+ +--------+ +--------+
| Node 70|--------------| Node 80|--------------| Node 90|
+--------+ +--------+ +--------+
|
|
+--------+
| Node 85|
|(Client)|
+--------+
Because the graph is not fully connected, when a node wants to send a message to another node, it may need to route it through the network. For instance, Node 10 can talk directly to nodes 20 and 40, but not to Node 70. In order to send a message to Node 70, it would first send it to Node 40 with instructions to pass it along to Node 70. Different overlay algorithms will have different connectivity graphs, but the general idea behind all of them is to allow any node in the graph to efficiently reach every other node within a small number of hops.
The RELOAD network is not only a messaging network. It is also a storage network. Records are stored under numeric addresses which occupy the same space as node identifiers. Peers are responsible for storing the data associated with some set of addresses as determined by their Node-ID. For instance, we might say that every peer is responsible for storing any data value which has an address less than or equal to its own Node-ID, but greater than the next lowest Node-ID. Thus, Node-20 would be responsible for storing values 11-20.
RELOAD also supports clients. These are nodes which have Node-IDs but do not participate in routing or storage. For instance, in the figure above Node 85 is a client. It can route to the rest of the RELOAD network via Node 80, but no other node will route through it and Node 90 is still responsible for all addresses between 81-90. We refer to non-client nodes as peers.
Other applications (for instance, SIP) can be defined on top of RELOAD and use these two basic RELOAD services to provide their own services.
RELOAD is fundamentally an overlay network. The following figure shows the layered RELOAD architecture.
Application
+-------+ +-------+
| SIP | | XMPP | ...
| Usage | | Usage |
+-------+ +-------+
------------------------------------ Messaging Service Boundary
+------------------+ +---------+
| Message |<--->| Storage |
| Transport | +---------+
+------------------+ ^
^ ^ |
| v v
| +-------------------+
| | Topology |
| | Plugin |
| +-------------------+
| ^
v v
+------------------+
| Forwarding & |
| Link Management |
+------------------+
------------------------------------ Overlay Link Service Boundary
+-------+ +------+
|TLS | |DTLS | ...
+-------+ +------+
The major components of RELOAD are:
To further clarify the roles of the various layers, this figure parallels the architecture with each layer's role from an overlay perspective and implementation layer in the internet:
| Internet Model |
Real | Equivalent | Reload
Internet | in Overlay | Architecture
-------------+-----------------+------------------------------------
| | +-------+ +-------+
| Application | | SIP | | XMPP | ...
| | | Usage | | Usage |
| | +-------+ +-------+
| | ----------------------------------
| |+------------------+ +---------+
| Transport || Message |<--->| Storage |
| || Transport | +---------+
| |+------------------+ ^
| | ^ ^ |
| | | v v
Application | | | +-------------------+
| (Routing) | | | Topology |
| | | | Plugin |
| | | +-------------------+
| | | ^
| | v v
| Network | +------------------+
| | | Forwarding & |
| | | Link Management |
| | +------------------+
| | ----------------------------------
Transport | Link | +-------+ +------+
| | |TLS | |DTLS | ...
| | +-------+ +------+
-------------+-----------------+------------------------------------
Network |
|
Link |
The top layer, called the Usage Layer, has application usages, such as the SIP Registration Usage [I-D.ietf-p2psip-sip], that use the abstract Message Transport Service provided by RELOAD. The goal of this layer is to implement application-specific usages of the generic overlay services provided by RELOAD. The usage defines how a specific application maps its data into something that can be stored in the overlay, where to store the data, how to secure the data, and finally how applications can retrieve and use the data.
The architecture diagram shows both a SIP usage and an XMPP usage. A single application may require multiple usages; for example a softphone application may also require a voicemail usage. An usage may define multiple kinds of data that are stored in the overlay and may also rely on kinds originally defined by other usages.
Because the security and storage policies for each kind are dictated by the usage defining the kind, the usages may be coupled with the Storage component to provide security policy enforcement and to implement appropriate storage strategies according to the needs of the usage. The exact implementation of such an interface is outside the scope of this specification.
The Message Transport component provides a generic message routing service for the overlay. The Message Transport layer is responsible for end-to-end message transactions, including retransmissions. Each peer is identified by its location in the overlay as determined by its Node-ID. A component that is a client of the Message Transport can perform two basic functions:
All usages rely on the Message Transport component to send and receive messages from peers. For instance, when a usage wants to store data, it does so by sending Store requests. Note that the Storage component and the Topology Plugin are themselves clients of the Message Transport, because they need to send and receive messages from other peers.
The Message Transport Service is similar to those described as providing "Key based routing" (KBR), although as RELOAD supports different overlay algorithms (including non-DHT overlay algorithms) that calculate keys in different ways, the actual interface must accept Resource Names rather than actual keys.
One of the major functions of RELOAD is to allow nodes to store data in the overlay and to retrieve data stored by other nodes or by themselves. The Storage component is responsible for processing data storage and retrieval messages. For instance, the Storage component might receive a Store request for a given resource from the Message Transport. It would then query the appropriate usage before storing the data value(s) in its local data store and sending a response to the Message Transport for delivery to the requesting node. Typically, these messages will come from other nodes, but depending on the overlay topology, a node might be responsible for storing data for itself as well, especially if the overlay is small.
A peer's Node-ID determines the set of resources that it will be responsible for storing. However, the exact mapping between these is determined by the overlay algorithm in use. The Storage component will only receive a Store request from the Message Transport if this peer is responsible for that Resource-ID. The Storage component is notified by the Topology Plugin when the Resource-IDs for which it is responsible change, and the Storage component is then responsible for migrating resources to other peers, as required.
RELOAD is explicitly designed to work with a variety of overlay algorithms. In order to facilitate this, the overlay algorithm implementation is provided by a Topology Plugin so that each overlay can select an appropriate overlay algorithm that relies on the common RELOAD core protocols and code.
The Topology Plugin is responsible for maintaining the overlay algorithm Routing Table, which is consulted by the Forwarding and Link Management Layer before routing a message. When connections are made or broken, the Forwarding and Link Management Layer notifies the Topology Plugin, which adjusts the routing table as appropriate. The Topology Plugin will also instruct the Forwarding and Link Management Layer to form new connections as dictated by the requirements of the overlay algorithm Topology. The Topology Plugin issues periodic update requests through Message Transport to maintain and update its Routing Table.
As peers enter and leave, resources may be stored on different peers, so the Topology Plugin also keeps track of which peers are responsible for which resources. As peers join and leave, the Topology Plugin instructs the Storage component to issue resource migration requests as appropriate, in order to ensure that other peers have whatever resources they are now responsible for. The Topology Plugin is also responsible for providing for redundant data storage to protect against loss of information in the event of a peer failure and to protect against compromised or subversive peers.
The Forwarding and Link Management Layer is responsible for getting a message to the next peer, as determined by the Topology Plugin. This Layer establishes and maintains the network connections as required by the Topology Plugin. This layer is also responsible for setting up connections to other peers through NATs and firewalls using ICE, and it can elect to forward traffic using relays for NAT and firewall traversal.
This layer provides a generic interface that allows the topology plugin to control the overlay and resource operations and messages. Since each overlay algorithm is defined and functions differently, we generically refer to the table of other peers that the overlay algorithm maintains and uses to route requests (neighbors) as a Routing Table. The Topology Plugin actually owns the Routing Table, and forwarding decisions are made by querying the Topology Plugin for the next hop for a particular Node-ID or Resource-ID. If this node is the destination of the message, the message is delivered to the Message Transport.
This layer also utilizes a framing header to encapsulate messages as they are forwarding along each hop. This header aids reliability congestion control, flow control, etc. It has meaning only in the context of that individual link.
The Forwarding and Link Management Layer sits on top of the Overlay Link Layer protocols that carry the actual traffic. This specification defines how to use DTLS and TLS protocols to carry RELOAD messages.
RELOAD's security model is based on each node having one or more public key certificates. In general, these certificates will be assigned by a central server which also assigns Node-IDs, although self-signed certificates can be used in closed networks. These credentials can be leveraged to provide communications security for RELOAD messages. RELOAD provides communications security at three levels:
These three levels of security work together to allow peers to verify the origin and correctness of data they receive from other peers, even in the face of malicious activity by other peers in the overlay. RELOAD also provides access control built on top of these communications security features. Because the peer responsible for storing a piece of data can validate the signature on the data being stored, the responsible peer can determine whether a given operation is permitted or not.
RELOAD also provides an optional shared secret based admission control feature using shared secrets and TLS-PSK. In order to form a TLS connection to any node in the overlay, a new node needs to know the shared overlay key, thus restricting access to authorized users only. This feature is used together with certificate-based access control, not as a replacement for it. It is typically used when self-signed certificates are being used but would generally not be used when the certificates were all signed by an enrollment server.
The remainder of this document is structured as follows.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].
We use the terminology and definitions from the Concepts and Terminology for Peer to Peer SIP [I-D.ietf-p2psip-concepts] draft extensively in this document. Other terms used in this document are defined inline when used and are also defined below for reference.
The term "maximum request lifetime" is the maximum time a request will wait for a response; it defaults to 15 seconds. The term "successor replacement hold-down time" is the amount of time to wait before starting replication when a new successor is found; it defaults to 30 seconds.
The most basic function of RELOAD is as a generic overlay network. Nodes need to be able to join the overlay, form connections to other nodes, and route messages through the overlay to nodes to which they are not directly connected. This section provides an overview of the mechanisms that perform these functions.
Every node in the RELOAD overlay is identified by a Node-ID. The Node-ID is used for three major purposes:
Each node has a certificate [RFC5280] containing a Node-ID, which is unique within an overlay instance.
The certificate serves multiple purposes:
If a user has more than one device, typically they would get one certificate for each device. This allows each device to act as a separate peer.
RELOAD supports multiple certificate issuance models. The first is based on a central enrollment process which allocates a unique name and Node-ID and puts them in a certificate for the user. All peers in a particular Overlay Instance have the enrollment server as a trust anchor and so can verify any other peer's certificate.
In some settings, a group of users want to set up an overlay network but are not concerned about attack by other users in the network. For instance, users on a LAN might want to set up a short term ad hoc network without going to the trouble of setting up an enrollment server. RELOAD supports the use of self-generated, self-signed certificates. When self-signed certificates are used, the node also generates its own Node-ID and username. The Node-ID is computed as a digest of the public key, to prevent Node-ID theft; however this model is still subject to a number of known attacks (most notably Sybil attacks [Sybil]) and can only be safely used in closed networks where users are mutually trusting.
The general principle here is that the security mechanisms (TLS and message signatures) are always used, even if the certificates are self-signed. This allows for a single set of code paths in the systems with the only difference being whether certificate verification is required to chain to a single root of trust.
RELOAD also provides an admission control system based on shared keys. In this model, the peers all share a single key which is used to authenticate the peer-to-peer connections via TLS-PSK/TLS-SRP.
RELOAD defines a single protocol that is used both as the peer protocol and as the client protocol for the overlay. This simplifies implementation, particularly for devices that may act in either role, and allows clients to inject messages directly into the overlay.
We use the term "peer" to identify a node in the overlay that routes messages for nodes other than those to which it is directly connected. Peers typically also have storage responsibilities. We use the term "client" to refer to nodes that do not have routing or storage responsibilities. When text applies to both peers and clients, we will simply refer such devices as "nodes."
RELOAD's client support allows nodes that are not participating in the overlay as peers to utilize the same implementation and to benefit from the same security mechanisms as the peers. Clients possess and use certificates that authorize the user to store data at certain locations in the overlay. The Node-ID in the certificate is used to identify the particular client as a member of the overlay and to authenticate its messages.
In RELOAD, unlike some other designs, clients are not a first-class concept. From the perspective of a peer, a client is simply a node which has not yet sent any Updates or Joins. It might never do so (if it's a client) or it might eventually do so (if it's just a node that's taking a long time to join). The routing and storage rules for RELOAD provide for correct behavior by peers regardless of whether other nodes attached to them are clients or peers. Of course, a client implementation must know that it intends to be a client, but this localizes complexity only to that node.
For more discussion of the motivation for RELOAD's client support, see Appendix Appendix C.
Clients may insert themselves in the overlay in two ways:
A node may act as a client simply because it does not have the resources or even an implementation of the topology plugin required to act as a peer in the overlay. In order to exchange RELOAD messages with a peer, a client must meet a minimum level of functionality. Such a client must:
A client speaks the same protocol as the peers, knows how to calculate Resource-IDs, and signs its requests in the same manner as peers. While a client does not necessarily require a full implementation of the overlay algorithm, calculating the Resource-ID requires an implementation of the appropriate algorithm for the overlay.
This section will discuss the requirements RELOAD's routing capabilities must meet, then describe the routing features in the protocol, and then provide a brief overview of how they are used. Appendix Appendix B discusses some alternative designs and the tradeoffs that would be necessary to support them.
RELOAD's routing capabilities must meet the following requirements:
RELOAD's routing provides three mechanisms designed to assist in meeting these needs:
The basic routing mechanism used by RELOAD is Symmetric Recursive. We will first describe symmetric recursive routing and then discuss its advantages in terms of the requirements discussed above.
Symmetric recursive routing requires that a message follow a path through the overlay to the destination without returning to the originating node: each peer forwards the message closer to its destination. The return path of the response is then the same path followed in reverse. For example, a message following a route from A to Z through B and X:
A B X Z
-------------------------------
---------->
Dest=Z
---------->
Via=A
Dest=Z
---------->
Via=A, B
Dest=Z
<----------
Dest=X, B, A
<----------
Dest=B, A
<----------
Dest=A
Note that the preceding Figure does not indicate whether A is a client or peer: A forwards its request to B and the response is returned to A in the same manner regardless of A's role in the overlay.
This figure shows use of full via-lists by intermediate peers B and X. However, if B and/or X are willing to store state, then they may elect to truncate the lists, save that information internally (keyed by the transaction id), and return the response message along the path from which it was received when the response is received. This option requires greater state to be stored on intermediate peers but saves a small amount of bandwidth and reduces the need for modifying the message en route. Selection of this mode of operation is a choice for the individual peer; the techniques are interoperable even on a single message. The figure below shows B using full via lists but X truncating them to X1 and saving the state internally.
A B X Z
-------------------------------
---------->
Dest=Z
---------->
Via=A
Dest=Z
---------->
Dest=Z, X1
<----------
Dest=X,X1
<----------
Dest=B, A
<----------
Dest=A
RELOAD also supports a basic Iterative routing mode (where the intermediate peers merely return a response indicating the next hop, but do not actually forward the message to that next hop themselves). Iterative routing is implemented using the RouteQuery method, which requests this behavior. Note that iterative routing is selected only by the initiating node.
In order to provide efficient routing, a peer needs to maintain a set of direct connections to other peers in the Overlay Instance. Due to the presence of NATs, these connections often cannot be formed directly. Instead, we use the Attach request to establish a connection. Attach uses ICE [RFC5245] to establish the connection. It is assumed that the reader is familiar with ICE.
Say that peer A wishes to form a direct connection to peer B. It gathers ICE candidates and packages them up in an Attach request which it sends to B through usual overlay routing procedures. B does its own candidate gathering and sends back a response with its candidates. A and B then do ICE connectivity checks on the candidate pairs. The result is a connection between A and B. At this point, A and B can add each other to their routing tables and send messages directly between themselves without going through other overlay peers.
There is one special case in which Attach cannot be used: when a peer is joining the overlay and is not connected to any peers. In order to support this case, some small number of "bootstrap nodes" typically need to be publicly accessible so that new peers can directly connect to them. Section 10 contains more detail on this.
In general, a peer needs to maintain connections to all of the peers near it in the Overlay Instance and to enough other peers to have efficient routing (the details depend on the specific overlay). If a peer cannot form a connection to some other peer, this isn't necessarily a disaster; overlays can route correctly even without fully connected links. However, a peer should try to maintain the specified link set and if it detects that it has fewer direct connections, should form more as required. This also implies that peers need to periodically verify that the connected peers are still alive and if not try to reform the connection or form an alternate one.
The Topology Plugin allows RELOAD to support a variety of overlay algorithms. This specification defines a DHT based on Chord [Chord], which is mandatory to implement, but the base RELOAD protocol is designed to support a variety of overlay algorithms.
RELOAD defines three methods for overlay maintenance: Join, Update, and Leave. However, the contents of those messages, when they are sent, and their precise semantics are specified by the actual overlay algorithm; RELOAD merely provides a framework of commonly-needed methods that provides uniformity of notation (and ease of debugging) for a variety of overlay algorithms.
When a new peer wishes to join the Overlay Instance, it must have a Node-ID that it is allowed to use and a set of credentials which match that Node-ID. When an enrollment server is used that Node-ID will be in the certificate the node received from the enrollment server. The details of the joining procedure are defined by the overlay algorithm, but the general steps for joining an Overlay Instance are:
The first thing the peer needs to do is to form a connection to some "bootstrap node". Because this is the first connection the peer makes, these nodes must have public IP addresses so that they can be connected to directly. Once a peer has connected to one or more bootstrap nodes, it can form connections in the usual way by routing Attach messages through the overlay to other nodes. Once a peer has connected to the overlay for the first time, it can cache the set of nodes it has connected to with public IP addresses for use as future bootstrap nodes.
Once a peer has connected to a bootstrap node, it then needs to take up its appropriate place in the overlay. This requires two major operations:
The second operation is performed by contacting the Admitting Peer (AP), the node which is currently responsible for that section of the overlay.
The details of this operation depend mostly on the overlay algorithm involved, but a typical case would be:
After this process is completed, JP is a full member of the Overlay Instance and can process Store/Fetch requests.
Note that the first node is a special case. When ordinary nodes cannot form connections to the bootstrap nodes, then they are not part of the overlay. However, the first node in the overlay can obviously not connect to other nodes. In order to support this case, potential first nodes (which must also serve as bootstrap nodes initially) must somehow be instructed (perhaps by configuration settings) that they are the entire overlay, rather than not part of it.
Note that clients do not perform either of these operations.
Previous sections addressed how RELOAD works once a node has connected. This section provides an overview of how users get connected to the overlay for the first time. RELOAD is designed so that users can start with the name of the overlay they wish to join and perhaps a username and password, and leverage that into having a working peer with minimal user intervention. This helps avoid the problems that have been experienced with conventional SIP clients where users are required to manually configure a large number of settings.
In the first phase of the process, the user starts out with the name of the overlay and uses this to download an initial set of overlay configuration parameters. The node does a DNS SRV lookup on the overlay name to get the address of a configuration server. It can then connect to this server with HTTPS to download a configuration document which contains the basic overlay configuration parameters as well as a set of bootstrap nodes which can be used to join the overlay.
If a node already has the valid configuration document that it received by some out of band method, this step can be skipped.
If the overlay is using centralized enrollment, then a user needs to acquire a certificate before joining the overlay. The certificate attests both to the user's name within the overlay and to the Node-IDs which they are permitted to operate. In that case, the configuration document will contain the address of an enrollment server which can be used to obtain such a certificate. The enrollment server may (and probably will) require some sort of username and password before issuing the certificate. The enrollment server's ability to restrict attackers' access to certificates in the overlay is one of the cornerstones of RELOAD's security.
RELOAD is not intended to be used alone, but rather as a substrate for other applications. These applications can use RELOAD for a variety of purposes:
This section provides an overview of these services.
RELOAD provides operations to Store and Fetch data. Each location in the Overlay Instance is referenced by a Resource-ID. However, each location may contain data elements corresponding to multiple kinds (e.g., certificate, SIP registration). Similarly, there may be multiple elements of a given kind, as shown below:
+--------------------------------+
| Resource-ID |
| |
| +------------+ +------------+ |
| | Kind 1 | | Kind 2 | |
| | | | | |
| | +--------+ | | +--------+ | |
| | | Value | | | | Value | | |
| | +--------+ | | +--------+ | |
| | | | | |
| | +--------+ | | +--------+ | |
| | | Value | | | | Value | | |
| | +--------+ | | +--------+ | |
| | | +------------+ |
| | +--------+ | |
| | | Value | | |
| | +--------+ | |
| +------------+ |
+--------------------------------+
Each kind is identified by a Kind-ID, which is a code point either assigned by IANA or allocated out of a private range. As part of the kind definition, protocol designers may define constraints, such as limits on size, on the values which may be stored. For many kinds, the set may be restricted to a single value; some sets may be allowed to contain multiple identical items while others may only have unique items. Note that a kind may be employed by multiple usages and new usages are encouraged to use previously defined kinds where possible. We define the following data models in this document, though other usages can define their own structures:
In order to protect stored data from tampering, by other nodes, each stored value is digitally signed by the node which created it. When a value is retrieved, the digital signature can be verified to detect tampering.
A major issue in peer-to-peer storage networks is minimizing the burden of becoming a peer, and in particular minimizing the amount of data which any peer is required to store for other nodes. RELOAD addresses this issue by only allowing any given node to store data at a small number of locations in the overlay, with those locations being determined by the node's certificate. When a peer uses a Store request to place data at a location authorized by its certificate, it signs that data with the private key that corresponds to its certificate. Then the peer responsible for storing the data is able to verify that the peer issuing the request is authorized to make that request. Each data kind defines the exact rules for determining what certificate is appropriate.
The most natural rule is that a certificate authorizes a user to store data keyed with their user name X. This rule is used for all the kinds defined in this specification. Thus, only a user with a certificate for "alice@example.org" could write to that location in the overlay. However, other usages can define any rules they choose, including publicly writable values.
The digital signature over the data serves two purposes. First, it allows the peer responsible for storing the data to verify that this Store is authorized. Second, it provides integrity for the data. The signature is saved along with the data value (or values) so that any reader can verify the integrity of the data. Of course, the responsible peer can "lose" the value but it cannot undetectably modify it.
The size requirements of the data being stored in the overlay are variable. For instance, a SIP AOR and voicemail differ widely in the storage size. RELOAD leaves it to the Usage and overlay configuration to limit size imbalance of various kinds.
Replication in P2P overlays can be used to provide:
A variety of schemes are used in P2P overlays to achieve some of these goals. Common techniques include replicating on neighbors of the responsible peer, randomly locating replicas around the overlay, or replicating along the path to the responsible peer.
The core RELOAD specification does not specify a particular replication strategy. Instead, the first level of replication strategies are determined by the overlay algorithm, which can base the replication strategy on its particular topology. For example, Chord places replicas on successor peers, which will take over responsibility should the responsible peer fail [Chord].
If additional replication is needed, for example if data persistence is particularly important for a particular usage, then that usage may specify additional replication, such as implementing random replications by inserting a different well known constant into the Resource Name used to store each replicated copy of the resource. Such replication strategies can be added independent of the underlying algorithm, and their usage can be determined based on the needs of the particular usage.
By itself, the distributed storage layer just provides infrastructure on which applications are built. In order to do anything useful, a usage must be defined. Each Usage needs to specify several things:
The kinds defined by a usage may also be applied to other usages. However, a need for different parameters, such as different size limits, would imply the need to create a new kind.
RELOAD does not currently define a generic service discovery algorithm as part of the base protocol, although a simplistic TURN-specific discovery mechanism is provided. A variety of service discovery algorithms can be implemented as extensions to the base protocol, such as the service discovery algorithm ReDIR [opendht-sigcomm05] or [I-D.maenpaa-p2psip-service-discovery].
There is no requirement that a RELOAD usage must use RELOAD's primitives for establishing its own communication if it already possesses its own means of establishing connections. For example, one could design a RELOAD-based resource discovery protocol which used HTTP to retrieve the actual data.
For more common situations, however, it is the overlay itself - rather than an external authority such as DNS - which is used to establish a connection. RELOAD provides connectivity to applications using the AppAttach method. For example, if a P2PSIP node wishes to establish a SIP dialog with another P2PSIP node, it will use AppAttach to establish a direct connection with the other node. This new connection is separate from the peer protocol connection. It is a dedicated UDP or TCP flow used only for the SIP dialog.
This section defines the basic protocols used to create, maintain, and use the RELOAD overlay network. We start by defining the basic concept of how message destinations are interpreted when routing messages. We then describe the symmetric recursive routing model, which is RELOAD's default routing algorithm. We then define the message structure and then finally define the messages used to join and maintain the overlay.
When a peer receives a message, it first examines the overlay, version, and other header fields to determine whether the message is one it can process. If any of these are incorrect (e.g., the message is for an overlay in which the peer does not participate) it is an error. The peer SHOULD generate an appropriate error but local policy can override this and cause the messages to be silently dropped.
Once the peer has determined that the message is correctly formatted, it examines the first entry on the destination list. There are three possible cases here:
These cases are handled as discussed below.
If the first entry on the destination list is an ID for which the node is responsible, there are several sub-cases to consider.
Note that this implies that in order to address a message to "the peer that controls region X", a sender sends to Resource-ID X, not Node-ID X.
If neither of the other three cases applies, then the peer MUST forward the message towards the first entry on the destination list. This means that it MUST select one of the peers to which it is connected and which is likely to be responsible for the first entry on the destination list. If the first entry on the destination list is in the peer's connection table, then it SHOULD forward the message to that peer directly. Otherwise, the peer consults the routing table to forward the message.
Any intermediate peer which forwards a RELOAD request MUST arrange that if it receives a response to that message the response can be routed back through the set of nodes through which the request passed. This may be arranged in one of two ways:
As an example of the first strategy, if node D receives a message from node C with via list (A, B), then D would forward to the next node (E) with via list (A, B, C). Now, if E wants to respond to the message, it reverses the via list to produce the destination list, resulting in (D, C, B, A). When D forwards the response to C, the destination list will contain (C, B, A).
As an example of the second strategy, if node D receives a message from node C with transaction ID X and via list (A, B), it could store (X, C) in its state database and forward the message with the via list unchanged. When D receives the response, it consults its state database for transaction id X, determines that the request came from C, and forwards the response to C.
Intermediate peers which modify the via list are not required to simply add entries. The only requirement is that the peer be able to reconstruct the correct destination list on the return route. RELOAD provides explicit support for this functionality in the form of private IDs, which can replace any number of via list entries. For instance, in the above example, Node D might send E a via list containing only the private ID (I). E would then use the destination list (D, I) to send its return message. When D processes this destination list, it would detect that I is a private ID, recover the via list (A, B, C), and reverse that to produce the correct destination list (C, B, A) before sending it to C. This feature is called List Compression. It MAY either be a compressed version of the original via list or an index into a state database containing the original via list.
No matter what mechanism for storing via list state is used, if an intermediate peer exits the overlay, then on the return trip the message cannot be forwarded and will be dropped. The ordinary timeout and retransmission mechanisms provide stability over this type of failure.
Note that if an intermediate peer retains per-transaction state instead of modifying the via list, it needs some mechanism for timing out that state, otherwise its state database will grow without bound. Whatever algorithm is used, unless a FORWARD_CRITICAL forwarding option or overlay configuration option explicitly indicates this state is not needed, the state MUST be maintained for at least the value of the overlay reliability timer (3 seconds) and MAY be kept longer. Future extension, such as [I-D.jiang-p2psip-relay], may define mechanisms for determining when this state does not need to be retained.
None of the above mechanisms are required for responses, since there is no need to ensure that subsequent requests follow the same path.
If the first entry in the destination list is a private id (e.g., a compressed via list), the peer MUST replace that entry with the original via list that it replaced and then re-examine the destination list to determine which of the above cases now applies.
This Section defines RELOAD's symmetric recursive routing algorithm, which is the default algorithm used by nodes to route messages through the overlay. All implementations MUST implement this routing algorithm. An overlay may be configured to use alternative routing algorithms, and alternative routing algorithms may be selected on a per-message basis.
In order to originate a message to a given Node-ID or Resource-ID, a node constructs an appropriate destination list. The simplest such destination list is a single entry containing the Node-ID or Resource-ID. The resulting message will use the normal overlay routing mechanisms to forward the message to that destination. The node can also construct a more complicated destination list for source routing.
Once the message is constructed, the node sends the message to some adjacent peer. If the first entry on the destination list is directly connected, then the message MUST be routed down that connection. Otherwise, the topology plugin MUST be consulted to determine the appropriate next hop.
Parallel searches for the resource are a common solution to improve reliability in the face of churn or of subversive peers. Parallel searches for usage-specified replicas are managed by the usage layer. However, a single request can also be routed through multiple adjacent peers, even when known to be sub-optimal, to improve reliability [vulnerabilities-acsac04]. Such parallel searches MAY be specified by the topology plugin.
Because messages may be lost in transit through the overlay, RELOAD incorporates an end-to-end reliability mechanism. When an originating node transmits a request it MUST set a 3 second timer. If a response has not been received when the timer fires, the request is retransmitted with the same transaction identifier. The request MAY be retransmitted up to 4 times (for a total of 5 messages). After the timer for the fifth transmission fires, the message SHALL be considered to have failed. Note that this retransmission procedure is not followed by intermediate nodes. They follow the hop-by-hop reliability procedure described in Section 5.6.3.
The above algorithm can result in multiple requests being delivered to a node. Receiving nodes MUST generate semantically equivalent responses to retransmissions of the same request (this can be determined by transaction id) if the request is received within the maximum request lifetime (15 seconds). For some requests (e.g., Fetch) this can be accomplished merely by processing the request again. For other requests, (e.g., Store) it may be necessary to maintain state for the duration of the request lifetime.
When a peer sends a response to a request using this routing algorithm, it MUST construct the destination list by reversing the order of the entries on the via list. This has the result that the response traverses the same peers as the request traversed, except in reverse order (symmetric routing).
RELOAD is a message-oriented request/response protocol. The messages are encoded using binary fields. All integers are represented in network byte order. The general philosophy behind the design was to use Type, Length, Value fields to allow for extensibility. However, for the parts of a structure that were required in all messages, we just define these in a fixed position, as adding a type and length for them is unnecessary and would simply increase bandwidth and introduces new potential for interoperability issues.
Each message has three parts, concatenated as shown below:
+-------------------------+ | Forwarding Header | +-------------------------+ | Message Contents | +-------------------------+ | Security Block | +-------------------------+
The contents of these parts are as follows:
The following sections describe the format of each part of the message.
The structures defined in this document are defined using a C-like syntax based on the presentation language used to define TLS. [RFC5246] Advantages of this style include:
Several idiosyncrasies of this language are worth noting.
For instance, "uint16 array<0..2^8-2>;" represents up to 254 bytes but only up to 127 values of two bytes (16 bits) each.
The following definitions are used throughout RELOAD and so are defined here. They also provide a convenient introduction to how to read the presentation language.
An enum represents an enumerated type. The values associated with each possibility are represented in parentheses and the maximum value is represented as a nameless value, for purposes of describing the width of the containing integral type. For instance, Boolean represents a true or false:
enum { false (0), true(1), (255)} Boolean;
A boolean value is either a 1 or a 0. The max value of 255 indicates this is represented as a single byte on the wire.
The NodeId, shown below, represents a single Node-ID.
typedef opaque NodeId[NodeIdLength];
A NodeId is a fixed-length structure represented as a series of bytes, with the most significant byte first. The length is set on a per-overlay basis within the range of 16-20 bytes (128 to 160 bits). (See Section 10.1 for how NodeIdLength is set.) Note: the use of "typedef" here is an extension to the TLS language, but its meaning should be relatively obvious. Note the [ size ] syntax defines a fixed length element that does not include the length of the element in the on the wire encoding.
A ResourceId, shown below, represents a single Resource-ID.
typedef opaque ResourceId<0..2^8-1>;
Like a NodeId, a ResourceId is an opaque string of bytes, but unlike NodeIds, ResourceIds are variable length, up to 254 bytes (2040 bits) in length. On the wire, each ResourceId is preceded by a single length byte (allowing lengths up to 255). Thus, the 3-byte value "FOO" would be encoded as: 03 46 4f 4f. Note the < range > syntax defines a variable length element that does include the length of the element in the on the wire encoding. The number of bytes to encode the length on the wire is derived by range; i.e., it is the minimum number of bytes which can encode the largest range value.
A more complicated example is IpAddressPort, which represents a network address and can be used to carry either an IPv6 or IPv4 address:
enum {reservedAddr(0), ipv4_address (1), ipv6_address (2),
(255)} AddressType;
struct {
uint32 addr;
uint16 port;
} IPv4AddrPort;
struct {
uint128 addr;
uint16 port;
} IPv6AddrPort;
struct {
AddressType type;
uint8 length;
select (type) {
case ipv4_address:
IPv4AddrPort v4addr_port;
case ipv6_address:
IPv6AddrPort v6addr_port;
/* This structure can be extended */
};
} IpAddressPort;
The first two fields in the structure are the same no matter what kind of address is being represented:
By having the type and the length appear at the beginning of the structure regardless of the kind of address being represented, an implementation which does not understand new address type X can still parse the IpAddressPort field and then discard it if it is not needed.
The rest of the IpAddressPort structure is either an IPv4AddrPort or an IPv6AddrPort. Both of these simply consist of an address represented as an integer and a 16-bit port. As an example, here is the wire representation of the IPv4 address "192.0.2.1" with port "6100".
01 ; type = IPv4
06 ; length = 6
c0 00 02 01 ; address = 192.0.2.1
17 d4 ; port = 6100
Unless a given structure that uses a select explicitly allows for unknown types in the select, any unknown type SHOULD be treated as an parsing error and the whole message discarded with no response.
The forwarding header is defined as a ForwardingHeader structure, as shown below.
struct {
uint32 relo_token;
uint32 overlay;
uint16 configuration_sequence;
uint8 version;
uint8 ttl;
uint32 fragment;
uint32 length;
uint64 transaction_id;
uint32 max_response_length;
uint16 via_list_length;
uint16 destination_list_length;
uint16 options_length;
Destination via_list[via_list_length];
Destination destination_list
[destination_list_length];
ForwardingOptions options[options_length];
} ForwardingHeader;
The contents of the structure are:
In order to be part of the overlay, a node MUST have a copy of the overlay configuration document. In order to allow for configuration document changes, each version of the configuration document has a sequence number which is monotonically increasing mod 65536. Because the sequence number may in principle wrap, greater than or less than are interpreted by modulo arithmetic as in TCP.
When a destination node receives a request, it MUST check that the configuration_sequence field is equal to its own configuration sequence number. If they do not match, it MUST generate an error, either Error_Config_Too_Old or Error_Config_Too_New. In addition, if the configuration file in the request is too old, it MUST generate a ConfigUpdate message to update the requesting node. This allows new configuration documents to propagate quickly throughout the system. The one exception to this rule is that if the configuration_sequence field is equal to 0xffff, and the message type is ConfigUpdate, then the message MUST be accepted regardless of the receiving node's configuration sequence number. Since 65535 is a special value, peers sending a new configuration when the configuration sequence is currently 65534 MUST set the configuration sequence number to 0 when they send out a new configuration.
The destination list and via lists are sequences of Destination values:
enum {reserved(0), node(1), resource(2), compressed(3),
/* 128-255 not allowed */ (255) }
DestinationType;
select (destination_type) {
case node:
NodeId node_id;
case resource:
ResourceId resource_id;
case compressed:
opaque compressed_id<0..2^8-1>;
/* This structure may be extended with new types */
} DestinationData;
struct {
DestinationType type;
uint8 length;
DestinationData destination_data;
} Destination;
struct {
uint16 compressed_id; /* top bit MUST be 1 */
} Destination;
If a destination structure has its first bit set to 1, then it is a 16 bit integer. If the first bit is not set, then it is a structure starting with DestinationType. If it is a 16 bit integer, it is treated as if it were a full structure with a DestinationType of compressed and a compressed_id that was 2 bytes long with the value of the 16 bit integer. When the destination structure is not a 16 bit integer, it is the TLV structure with the following contents:
A DestinationData can be one of three types:
One possible encoding of the 16 bit integer version as an opaque identifier is to encode an index into a connection table. To avoid misrouting responses in the event a response is delayed and the connection table entry has changed, the identifier SHOULD be split between an index and a generation counter for that index. At startup, the generation counters should be initialized to random values. An implementation could use 12 bits for the connection table index and 3 bits for the generation counter. (Note that this does not suggest a 4096 entry connection table for every node, only the ability to encode for a larger connection table.) When a connection table slot is used for a new connection, the generation counter is incremented (with wrapping). Connection table slots are used on a rotating basis to maximize the time interval between uses of the same slot for different connections. When routing a message to an entry in the destination list encoding a connection table entry, the node confirms that the generation counter matches the current generation counter of that index before forwarding the message. If it does not match, the message is silently dropped.
The Forwarding header can be extended with forwarding header options, which are a series of ForwardingOptions structures:
enum { reservedForwarding(0), (255) }
ForwardingOptionsType;
struct {
ForwardingOptionsType type;
uint8 flags;
uint16 length;
select (type) {
/* This type may be extended */
} option;
} ForwardingOption;
Each ForwardingOption consists of the following values:
The second major part of a RELOAD message is the contents part, which is defined by MessageContents:
enum { reservedMessagesExtension(0), (2^16-1) } MessageExtensionType;
struct {
MessageExtensionType type;
Boolean critical;
opaque extension_contents<0..2^32-1>;
} MessageExtension;
struct {
uint16 message_code;
opaque message_body<0..2^32-1>;
MessageExtensions extensions<0..2^32-1>;
} MessageContents;
The contents of this structure are as follows:
All extensions have the following form:
A peer processing a request returns its status in the message_code field. If the request was a success, then the message code is the response code that matches the request (i.e., the next code up). The response payload is then as defined in the request/response descriptions.
If the request has failed, then the message code is set to 0xffff (error) and the payload MUST be an error_response PDU, as shown below.
When the message code is 0xffff, the payload MUST be an ErrorResponse.
public struct {
uint16 error_code;
opaque error_info<0..2^16-1>;
} ErrorResponse;
The contents of this structure are as follows:
The following error code values are defined. The numeric values for these are defined in Section 13.9.
The third part of a RELOAD message is the security block. The security block is represented by a SecurityBlock structure:
enum { x509(0), (255) } certificate_type;
struct {
certificate_type type;
opaque certificate<0..2^16-1>;
} GenericCertificate;
struct {
GenericCertificate certificates<0..2^16-1>;
Signature signature;
} SecurityBlock;
The contents of this structure are:
The certificates bucket SHOULD contain all the certificates necessary to verify every signature in both the message and the internal message objects. This is the only location in the message which contains certificates, thus allowing for only a single copy of each certificate to be sent. In systems which have some alternate certificate distribution mechanism, some certificates MAY be omitted. However, implementors should note that this creates the possibility that messages may not be immediately verifiable because certificates must first be retrieved.
Each certificate is represented by a GenericCertificate structure, which has the following contents:
The signature is computed over the payload and parts of the forwarding header. The payload, in case of a Store, may contain an additional signature computed over a StoreReq structure. All signatures are formatted using the Signature element. This element is also used in other contexts where signatures are needed. The input structure to the signature computation varies depending on the data element being signed.
enum { reservedSignerIdentity(0),
cert_hash(1), (255)} SignerIdentityType;
struct {
select (identity_type) {
case cert_hash;
HashAlgorithm hash_alg; // From TLS
opaque certificate_hash<0..2^8-1>;
/* This structure may be extended with new types if necessary*/
};
} SignerIdentityValue;
struct {
SignerIdentityType identity_type;
uint16 length;
SignerIdentityValue identity[SignerIdentity.length];
} SignerIdentity;
struct {
SignatureAndHashAlgorithm algorithm; // From TLS
SignerIdentity identity;
opaque signature_value<0..2^16-1>;
} Signature;
The signature construct contains the following values:
The only currently permitted identity format is a hash of the signer's certificate. The hash_alg field is used to indicate the algorithm used to produce the hash. The certificate_hash contains the hash of the certificate object (i.e., the DER-encoded certificate). The SignerIdentity structure is typed purely to allow for future (unanticipated) extensibility.
For signatures over messages the input to the signature is computed over:
where overlay and transaction_id come from the forwarding header and || indicates concatenation.
The input to signatures over data values is different, and is described in Section 6.1.
All RELOAD messages MUST be signed. Upon receipt, the receiving node MUST verify the signature and the authorizing certificate. This check provides a minimal level of assurance that the sending node is a valid part of the overlay as well as cryptographic authentication of the sending node. In addition, responses MUST be checked as follows:
The second condition serves as a primitive check for responses from wildly wrong nodes but is not a complete check. Note that in periods of churn, it is possible for the requesting node to obtain a closer neighbor while the request is outstanding. This will cause the response to be rejected and the request to be retransmitted.
In addition, some methods (especially Store) have additional authentication requirements, which are described in the sections covering those methods.
As discussed in previous sections, RELOAD does not itself implement any overlay topology. Rather, it relies on Topology Plugins, which allow a variety of overlay algorithms to be used while maintaining the same RELOAD core. This section describes the requirements for new topology plugins and the methods that RELOAD provides for overlay topology maintenance.
When specifying a new overlay algorithm, at least the following need to be described:
All overlay algorithms MUST specify maintenance procedures that send Updates to clients and peers that have established connections to the peer responsible for a particular ID when the responsibility for that ID changes. Because tracking this information is difficult, overlay algorithms MAY simply specify that an Update is sent to all members of the Connection Table whenever the range of IDs for which the peer is responsible changes.
This section describes the methods that topology plugins use to join, leave, and maintain the overlay.
A new peer (but one that already has credentials) uses the JoinReq message to join the overlay. The JoinReq is sent to the responsible peer depending on the routing mechanism described in the topology plugin. This notifies the responsible peer that the new peer is taking over some of the overlay and it needs to synchronize its state.
struct {
NodeId joining_peer_id;
opaque overlay_specific_data<0..2^16-1>;
} JoinReq;
The minimal JoinReq contains only the Node-ID which the sending peer wishes to assume. Overlay algorithms MAY specify other data to appear in this request. Receivers of the JoinReq MUST verify that the joining_peer_id field matches the Node-ID used to sign the message and if not MUST reject the message with an Error_Forbidden error.
If the request succeeds, the responding peer responds with a JoinAns message, as defined below:
struct {
opaque overlay_specific_data<0..2^16-1>;
} JoinAns;
If the request succeeds, the responding peer MUST follow up by executing the right sequence of Stores and Updates to transfer the appropriate section of the overlay space to the joining peer. In addition, overlay algorithms MAY define data to appear in the response payload that provides additional info.
In general, nodes which cannot form connections SHOULD report an error. However, implementations MUST provide some mechanism whereby nodes can determine that they are potentially the first node and take responsibility for the overlay. This specification does not mandate any particular mechanism, but a configuration flag or setting seems appropriate.
The LeaveReq message is used to indicate that a node is exiting the overlay. A node SHOULD send this message to each peer with which it is directly connected prior to exiting the overlay.
struct {
NodeId leaving_peer_id;
opaque overlay_specific_data<0..2^16-1>;
} LeaveReq;
LeaveReq contains only the Node-ID of the leaving peer. Overlay algorithms MAY specify other data to appear in this request. Receivers of the LeaveReq MUST verify that the leaving_peer_id field matches the Node-ID used to sign the message and if not MUST reject the message with an Error_Forbidden error.
Upon receiving a Leave request, a peer MUST update its own routing table, and send the appropriate Store/Update sequences to re-stabilize the overlay.
Update is the primary overlay-specific maintenance message. It is used by the sender to notify the recipient of the sender's view of the current state of the overlay (its routing state), and it is up to the recipient to take whatever actions are appropriate to deal with the state change. In general, peers send Update messages to all their adjacencies whenever they detect a topology shift.
When a peer receives an Attach request with the send_update flag set to "true" Section 5.4.2.4, it MUST send an Update message back to the sender of the Attach request after the completion of the corresponding ICE check and TLS connection. Note that the sender of a such Attach request may not have joined the overlay yet.
When a peer detects through an Update that it is no longer responsible for any data value it is storing, it MUST attempt to Store a copy to the correct node unless it knows the newly responsible node already has a copy of the data. This prevents data loss during large-scale topology shifts such as the merging of partitioned overlays.
The contents of the UpdateReq message are completely overlay-specific. The UpdateAns response is expected to be either success or an error.
The RouteQuery request allows the sender to ask a peer where they would route a message directed to a given destination. In other words, a RouteQuery for a destination X requests the Node-ID for the node that the receiving peer would next route to in order to get to X. A RouteQuery can also request that the receiving peer initiate an Update request to transfer the receiving peer's routing table.
One important use of the RouteQuery request is to support iterative routing. The sender selects one of the peers in its routing table and sends it a RouteQuery message with the destination_object set to the Node-ID or Resource-ID it wishes to route to. The receiving peer responds with information about the peers to which the request would be routed. The sending peer MAY then use the Attach method to attach to that peer(s), and repeat the RouteQuery. Eventually, the sender gets a response from a peer that is closest to the identifier in the destination_object as determined by the topology plugin. At that point, the sender can send messages directly to that peer.
A RouteQueryReq message indicates the peer or resource that the requesting node is interested in. It also contains a "send_update" option allowing the requesting node to request a full copy of the other peer's routing table.
struct {
Boolean send_update;
Destination destination;
opaque overlay_specific_data<0..2^16-1>;
} RouteQueryReq;
The contents of the RouteQueryReq message are as follows:
A response to a successful RouteQueryReq request is a RouteQueryAns message. This is completely overlay specific.
Probe provides primitive "exploration" services: it allows node to determine which resources another node is responsible for; and it allows some discovery services using multicast, anycast, or broadcast. A probe can be addressed to a specific Node-ID, or the peer controlling a given location (by using a Resource-ID). In either case, the target Node-IDs respond with a simple response containing some status information.
The ProbeReq message contains a list (potentially empty) of the pieces of status information that the requester would like the responder to provide.
enum { reservedProbeInformation(0), responsible_set(1),
num_resources(2), uptime(3), (255)}
ProbeInformationType;
struct {
ProbeInformationType requested_info<0..2^8-1>;
} ProbeReq
The currently defined values for ProbeInformation are:
A successful ProbeAns response contains the information elements requested by the peer.
struct {
select (type) {
case responsible_set:
uint32 responsible_ppb;
case num_resources:
uint32 num_resources;
case uptime:
uint32 uptime;
/* This type may be extended */
};
} ProbeInformationData;
struct {
ProbeInformationType type;
uint8 length;
ProbeInformationData value;
} ProbeInformation;
struct {
ProbeInformation probe_info<0..2^16-1>;
} ProbeAns;
A ProbeAns message contains a sequence of ProbeInformation structures. Each has a "length" indicating the length of the following value field. This structure allows for unknown option types.
Each of the current possible Probe information types is a 32-bit unsigned integer. For type "responsible_ppb", it is the fraction of the overlay for which the peer is responsible in parts per billion. For type "num_resources", it is the number of resources the peer is storing. For the type "uptime" it is the number of seconds the peer has been up.
The responding peer SHOULD include any values that the requesting node requested and that it recognizes. They SHOULD be returned in the requested order. Any other values MUST NOT be returned.
Each node maintains connections to a set of other nodes defined by the topology plugin. This section defines the methods RELOAD uses to form and maintain connections between nodes in the overlay. Three methods are defined:
A node sends an Attach request when it wishes to establish a direct TCP or UDP connection to another node for the purpose of sending RELOAD messages.
As described in Section 5.1, an Attach may be routed to either a Node-ID or to a Resource-ID. An Attach routed to a specific Node-ID will fail if that node is not reached. An Attach routed to a Resource-ID will establish a connection with the peer currently responsible for that Resource-ID, which may be useful in establishing a direct connection to the responsible peer for use with frequent or large resource updates.
An Attach in and of itself does not result in updating the routing table of either node. That function is performed by Updates. If node A has Attached to node B, but not received any Updates from B, it MAY route messages which are directly addressed to B through that channel but MUST NOT route messages through B to other peers via that channel. The process of Attaching is separate from the process of becoming a peer (using Join and Update), to prevent half-open states where a node has started to form connections but is not really ready to act as a peer. Thus, clients (unlike peers) can simply Attach without sending Join or Update.
An Attach request message contains the requesting node ICE connection parameters formatted into a binary structure.
enum { reservedOverlayLink(0), DTLS-UDP-SR(1),
DTLS-UDP-SR-NO-ICE(3), TLS-TCP-FH-NO-ICE(4),
(255) } OverlayLinkType;
enum { reservedCand(0), host(1), srflx(2), prflx(3), relay(4),
(255) } CandType;
struct {
opaque name<0..2^16-1>;
opaque value<0..2^16-1>;
} IceExtension;
struct {
IpAddressPort addr_port;
OverlayLinkType overlay_link;
opaque foundation<0..255>;
uint32 priority;
CandType type;
select (type){
case host:
; /* Nothing */
case srflx:
case prflx:
case relay:
IpAddressPort rel_addr_port;
};
IceExtension extensions<0..2^16-1>;
} IceCandidate;
struct {
opaque ufrag<0..2^8-1>;
opaque password<0..2^8-1>;
opaque role<0..2^8-1>;
IceCandidate candidates<0..2^16-1>;
Boolean send_update;
} AttachReqAns;
The values contained in AttachReqAns are:
Each ICE candidate is represented as an IceCandidate structure, which is a direct translation of the information from the ICE string structures, with the exception of the component ID. Since there is only one component, it is always 1, and thus left out of the PDU. The remaining values are specified as follows:
These values should be generated using the procedures described in Section 5.5.1.3.
If a peer receives an Attach request, it MUST determine how to process the request as follows:
If the peer is overloaded or detects some other kind of error, it MAY generate an error instead of an AttachReqAns.
When a peer receives an Attach response, it SHOULD parse the response and begin its own ICE checks.
This section describes the profile of ICE that is used with RELOAD. RELOAD implementations MUST implement full ICE.
In ICE as defined by [RFC5245], SDP is used to carry the ICE parameters. In RELOAD, this function is performed by a binary encoding in the Attach method. This encoding is more restricted than the SDP encoding because the RELOAD environment is simpler:
An agent follows the ICE specification as described in [RFC5245] with the changes and additional procedures described in the subsections below.
ICE relies on the node having one or more STUN servers to use. In conventional ICE, it is assumed that nodes are configured with one or more STUN servers through some out of band mechanism. This is still possible in RELOAD but RELOAD also learns STUN servers as it connects to other peers. Because all RELOAD peers implement ICE and use STUN keepalives, every peer is a capable of responding to STUN Binding requests [RFC5389]. Accordingly, any peer that a node knows about can be used like a STUN server -- though of course it may be behind a NAT.
A peer on a well-provisioned wide-area overlay will be configured with one or more bootstrap nodes. These nodes make an initial list of STUN servers. However, as the peer forms connections with additional peers, it builds more peers it can use like STUN servers.
Because complicated NAT topologies are possible, a peer may need more than one STUN server. Specifically, a peer that is behind a single NAT will typically observe only two IP addresses in its STUN checks: its local address and its server reflexive address from a STUN server outside its NAT. However, if there are more NATs involved, it may learn additional server reflexive addresses (which vary based on where in the topology the STUN server is). To maximize the chance of achieving a direct connection, a peer SHOULD group other peers by the peer-reflexive addresses it discovers through them. It SHOULD then select one peer from each group to use as a STUN server for future connections.
Only peers to which the peer currently has connections may be used. If the connection to that host is lost, it MUST be removed from the list of stun servers and a new server from the same group MUST be selected unless there are no others servers in the group in which case some other peer MAY be used.
When a node wishes to establish a connection for the purposes of RELOAD signaling or application signaling, it follows the process of gathering candidates as described in Section 4 of ICE [RFC5245]. RELOAD utilizes a single component. Consequently, gathering for these "streams" requires a single component. In the case where a node has not yet found a TURN server, the agent would not include a relayed candidate.
The ICE specification assumes that an ICE agent is configured with, or somehow knows of, TURN and STUN servers. RELOAD provides a way for an agent to learn these by querying the overlay, as described in Section 5.5.1.4 and Section 8.
The default candidate selection described in Section 4.1.4 of ICE is ignored; defaults are not signaled or utilized by RELOAD.
An alternative to using the full ICE supported by the Attach request is to use No-ICE mechanism by providing candidates with "No-ICE" Overlay Link protocols. Configuration for the overlay indicates whether or not these Overlay Link protocols can be used. An overlay MUST be either all ICE or all No-ICE.
No-ICE will not work in all of the scenarios where ICE would work, but in some cases, particularly those with no NATs or firewalls, it will work. Therefore it is RECOMMENDED that full ICE be used even for a node that has a public, unfiltered IP address, to take advantage of STUN connectivity checks, etc.
At the time of writing, UDP is the only transport protocol specified to work with ICE. However, standardization of additional protocols for use with ICE is expected, including TCP and datagram-oriented protocols such as SCTP and DCCP. In particular, UDP encapsulations for SCTP and DCCP are expected to be standardized in the near future, greatly expanding the available Overlay Link protocols available for RELOAD. When additional protocols are available, the following prioritization is RECOMMENDED:
Section 4.3 of ICE describes procedures for encoding the SDP for conveying RELOAD candidates. Instead of actually encoding an SDP, the candidate information (IP address and port and transport protocol, priority, foundation, type and related address) is carried within the attributes of the Attach request or its response. Similarly, the username fragment and password are carried in the Attach message or its response. Section 5.5.1 describes the detailed attribute encoding for Attach. The Attach request and its response do not contain any default candidates or the ice-lite attribute, as these features of ICE are not used by RELOAD.
Since the Attach request contains the candidate information and short term credentials, it is considered as an offer for a single media stream that happens to be encoded in a format different than SDP, but is otherwise considered a valid offer for the purposes of following the ICE specification. Similarly, the Attach response is considered a valid answer for the purposes of following the ICE specification.
An agent MUST skip the verification procedures in Section 5.1 and 6.1 of ICE. Since RELOAD requires full ICE from all agents, this check is not required.
The roles of controlling and controlled as described in Section 5.2 of ICE are still utilized with RELOAD. However, the offerer (the entity sending the Attach request) will always be controlling, and the answerer (the entity sending the Attach response) will always be controlled. The connectivity checks MUST still contain the ICE-CONTROLLED and ICE-CONTROLLING attributes, however, even though the role reversal capability for which they are defined will never be needed with RELOAD. This is to allow for a common codebase between ICE for RELOAD and ICE for SDP.
When neither side has provided an No-ICE candidate, connectivity checks and nominations are used as in regular ICE.
The processes of forming check lists in Section 5.7 of ICE, scheduling checks in Section 5.8, and checking connectivity checks in Section 7 are used with RELOAD without change.
The procedures in Section 8 of ICE are followed to conclude ICE, with the following exceptions:
STUN MUST be utilized for the keepalives described in Section 10 of ICE.
No-ICE is selected when either side has provided "no ICE" Overlay Link candidates. STUN is not used for connectivity checks when doing No-ICE; instead the DTLS or TLS handshake (or similar security layer of future overlay link protocols) forms the connectivity check. The certificate exchanged during the (D)TLS handshake must match the node that sent the AttachReqAns and if it does not, the connection MUST be closed.
An agent MUST NOT send a subsequent offer or answer. Thus, the procedures in Section 9 of ICE MUST be ignored.
The procedures of Section 11 of ICE apply to RELOAD as well. However, in this case, the "media" takes the form of application layer protocols (RELOAD) over TLS or DTLS. Consequently, once ICE processing completes, the agent will begin TLS or DTLS procedures to establish a secure connection. The node which sent the Attach request MUST be the TLS server. The other node MUST be the TLS client. The server MUST request TLS client authentication. The nodes MUST verify that the certificate presented in the handshake matches the identity of the other peer as found in the Attach message. Once the TLS or DTLS signaling is complete, the application protocol is free to use the connection.
The concept of a previous selected pair for a component does not apply to RELOAD, since ICE restarts are not possible with RELOAD.
An agent MUST be prepared to receive packets for the application protocol (TLS or DTLS carrying RELOAD, SIP or anything else) at any time. The jitter and RTP considerations in Section 11 of ICE do not apply to RELOAD.
A node sends an AppAttach request when it wishes to establish a direct connection to another node for the purposes of sending application layer messages. AppAttach is nearly identical to Attach, except for the purpose of the connection: it is used to transport non-RELOAD "media". A separate request is used to avoid implementor confusion between the two methods (this was found to be a real problem with initial implementations). The AppAttach request and its response contain an application attribute, which indicates what protocol is to be run over the connection.
An AppAttachReq message contains the requesting node's ICE connection parameters formatted into a binary structure.
struct {
opaque ufrag<0..2^8-1>;
opaque password<0..2^8-1>;
uint16 application;
opaque role<0..2^8-1>;
IceCandidate candidates<0..2^16-1>;
} AppAttachReq;
The values contained in AppAttachReq and AppAttachAns are:
The application using connection set up with this request is responsible for providing sufficiently frequent keep traffic for NAT and Firewall keep alive and for deciding when to close the connection.
If a peer receives an AppAttach request, it SHOULD process the request and generate its own response with a AppAttachAns. It should then begin ICE checks. When a peer receives an AppAttach response, it SHOULD parse the response and begin its own ICE checks. If the application ID is not supported, the peer MUST reply with an Error_Not_Found error.
struct {
opaque ufrag<0..2^8-1>;
opaque password<0..2^8-1>;
uint16 application;
opaque role<0..2^8-1>;
IceCandidate candidates<0..2^16-1>;
} AppAttachAns;
The meaning of the fields is the same as in the AppAttachReq.
Ping is used to test connectivity along a path. A ping can be addressed to a specific Node-ID, to the peer controlling a given location (by using a resource ID), or to the broadcast Node-ID (2^128-1).
struct {
opaque<0..2^16-1> padding;
} PingReq
The Ping request is empty of meaningful contents. However, it may contain up to 65535 bytes of padding to facilitate the discovery of overlay maximum packet sizes.
A successful PingAns response contains the information elements requested by the peer.
struct {
uint64 response_id;
uint64 time;
} PingAns;
A PingAns message contains the following elements:
The ConfigUpdate method is used to push updated configuration data across the overlay. Whenever a node detects that another node has old configuration data, it MUST generate a ConfigUpdate request. The ConfigUpdate request allows updating of two kinds of data: the configuration data (Section 5.3.2.1) and kind information (Section 6.4.1.1).
enum { reservedConfigUpdate(0), config(1), kind(2), (255) }
ConfigUpdateType;
typedef uint32 KindId;
typedef opaque KindDescription<0..2^16-1>;
struct {
ConfigUpdateType type;
uint32 length;
select (type) {
case config:
opaque config_data<0..2^24-1>;
case kind:
KindDescription kinds<0..2^24-1>;
/* This structure may be extended with new types*/
};
} ConfigUpdateReq;
The ConfigUpdateReq message contains the following elements:
struct {
} ConfigUpdateAns
If the ConfigUpdateReq is of type "config" it MUST only be processed if all the following are true:
Otherwise appropriate errors MUST be generated.
If the ConfigUpdateReq is of type "kind" it MUST only be processed if it is correctly digitally signed by an acceptable kind signer as specified in the configuration file. Details on kind-signer field in the configuration file is described in Section 10.1. In addition, if the kind update conflicts with an existing known kind (i.e., it is signed by a different signer), then it should be rejected with "Error_Forbidden". This should not happen in correctly functioning overlays.
If the update is acceptable, then the node MUST reconfigure itself to match the new information. This may include adding permissions for new kinds, deleting old kinds, or even, in extreme circumstances, exiting and reentering the overlay, if, for instance, the DHT algorithm has changed.
The response for ConfigUpdate is empty.
RELOAD can use multiple Overlay Link protocols to send its messages. Because ICE is used to establish connections (see Section 5.5.1.3), RELOAD nodes are able to detect which Overlay Link protocols are offered by other nodes and establish connections between them. Any link protocol needs to be able to establish a secure, authenticated connection and to provide data origin authentication and message integrity for individual data elements. RELOAD currently supports three Overlay Link protocols:
Note that although UDP does not properly have "connections", both TLS and DTLS have a handshake which establishes a similar, stateful association, and we simply refer to these as "connections" for the purposes of this document.
If a peer receives a message that is larger than value of max-message-size defined in the overlay configuration, the peer SHOULD send an Error_Message_Too_Large error and then close the TLS or DTLS session from which the message was received. Note that this error can be sent and the session closed before receiving the complete message. If the forwarding header is larger than the max-message-size, the receiver SHOULD close the TLS or DTLS session without sending an error.
The Framing Header (FH) is used to frame messages and provide timing when used on a reliable stream-based transport protocol. Simple Reliability (SR) makes use of the FH to provide congestion control and semi-reliability when using unreliable message-oriented transport protocols. We will first define each of these algorithms, then define overlay link protocols that use them.
Note: We expect future Overlay Link protocols to define replacements for all components of these protocols, including the framing header. These protocols have been chosen for simplicity of implementation and reasonable performance.
Note to implementers: There are inherent tradeoffs in utilizing short timeouts to determine when a link has failed. To balance the tradeoffs, an implementation should be able to quickly act to remove entries from the routing table when there is reason to suspect the link has failed. For example, in a Chord-derived overlay algorithm, a closer finger table entry could be substituted for an entry in the finger table that has experienced a timeout. That entry can be restored if it proves to resume functioning, or replaced at some point in the future if necessary. End-to-end retransmissions will handle any lost messages, but only if the failing entries do not remain in the finger table for subsequent retransmissions.
The only currently defined overlay link protocols are TLS and DTLS. It is possible to define new link-layer protocols and apply them to a new overlay using the "overlay-link-protocol" configuration directive (see Section 10.1.). However, any new protocols MUST meet the following requirements.
Any new overlay protocol MUST be defined via RFC 5226 Standards Action; see Section 13.11.
In a Host Identity Protocol Based Overlay Networking Environment (HIP BONE) [I-D.ietf-hip-bone] HIP [RFC5201] provides connection management (e.g., NAT traversal and mobility) and security for the overlay network. The P2PSIP Working Group has expressed interest in supporting a HIP-based link protocol. Such support would require specifying such details as:
[I-D.ietf-hip-reload-instance] documents work in progress on using RELOAD with the HIP BONE.
The ICE-TCP draft [I-D.ietf-mmusic-ice-tcp] should allow TCP to be supported as an Overlay Link protocol that can be added using ICE.
Modern message-oriented transports offer high performance, good congestion control, and avoid head of line blocking in case of lost data. These characteristics make them preferable as underlying transport protocols for RELOAD links. SCTP without message ordering and DCCP are two examples of such protocols. However, currently they are not well-supported by commonly available NATs, and specifications for ICE session establishment are not available.
As of the time of this writing, there is significant interest in the IETF community in tunneling other transports over UDP, motivated by the situation that UDP is well-supported by modern NAT hardware, and similar performance can be achieved to native implementation. Currently SCTP, DCCP, and a generic tunneling extension are being proposed for message-oriented protocols. Baset et al. have proposed tunneling TCP over UDP for similar reasons [I-D.baset-tsvwg-tcp-over-udp]. Once ICE traversal has been specified for these tunneled protocols, they should be straightforward to support as overlay link protocols.
In order to support unreliable links and to allow for quick detection of link failures when using reliable end-to-end transports, each message is wrapped in a very simple framing layer (FramedMessage) which is only used for each hop. This layer contains a sequence number which can then be used for ACKs. The same header is used for both reliable and unreliable transports for simplicity of implementation.
The definition of FramedMessage is:
enum { data(128), ack(129), (255)} FramedMessageType;
struct {
FramedMessageType type;
select (type) {
case data:
uint32 sequence;
opaque message<0..2^24-1>;
case ack:
uint32 ack_sequence;
uint32 received;
};
} FramedMessage;
The type field of the PDU is set to indicate whether the message is data or an acknowledgement.
If the message is of type "data", then the remainder of the PDU is as follows:
Each connection has it own sequence number space. Initially the value is zero and it increments by exactly one for each message sent over that connection.
When the receiver receives a message, it SHOULD immediately send an ACK message. The receiver MUST keep track of the 32 most recent sequence numbers received on this association in order to generate the appropriate ack.
If the PDU is of type "ack", the contents are as follows:
The received field bits in the ACK provide a high degree of redundancy so that the sender can figure out which packets the receiver has received and can then estimate packet loss rates. If the sender also keeps track of the time at which recent sequence numbers have been sent, the RTT can be estimated.
When RELOAD is carried over DTLS or another unreliable link protocol, it needs to be used with a reliability and congestion control mechanism, which is provided on a hop-by-hop basis. The basic principle is that each message, regardless of whether or not it carries a request or response, will get an ACK and be reliably retransmitted. The receiver's job is very simple, limited to just sending ACKs. All the complexity is at the sender side. This allows the sending implementation to trade off performance versus implementation complexity without affecting the wire protocol.
Because the receiver's role is limited to providing packet acknowledgements, a wide variety of congestion control algorithms can be implemented on the sender side while using the same basic wire protocol. In general, senders MAY implement any rate control scheme of their choice, provided that it is REQUIRED to be no more aggressive then TFRC[RFC5348].
The following section describes a simple, inefficient, scheme that complies with this requirement. Another alternative would be TFRC-SP [RFC4828] and use the received bitmask to allow the sender to compute packet loss event rates.
A node SHOULD retransmit a message if it has not received an ACK after an interval of RTO ("Retransmission TimeOut"). The node MUST double the time to wait after each retransmission. In each retransmission, the sequence number is incremented.
The RTO is an estimate of the round-trip time (RTT). Implementations can use a static value for RTO or a dynamic estimate which will result in better performance. For implementations that use a static value, the default value for RTO is 500 ms. Nodes MAY use smaller values of RTO if it is known that all nodes are within the local network. The default RTO MAY be chosen larger, and this is RECOMMENDED if it is known in advance (such as on high latency access links) that the round-trip time is larger.
Implementations that use a dynamic estimate to compute the RTO MUST use the algorithm described in RFC 2988[RFC2988], with the exception that the value of RTO SHOULD NOT be rounded up to the nearest second but instead rounded up to the nearest millisecond. The RTT of a successful STUN transaction from the ICE stage is used as the initial measurement for formula 2.2 of RFC 2988. The sender keeps track of the time each message was sent for all recently sent messages. Any time an ACK is received, the sender can compute the RTT for that message by looking at the time the ACK was received and the time when the message was sent. This is used as a subsequent RTT measurement for formula 2.3 of RFC 2988 to update the RTO estimate. (Note that because retransmissions receive new sequence numbers, all received ACKs are used.)
The value for RTO is calculated separately for each DTLS session.
Retransmissions continue until a response is received, or until a total of 5 requests have been sent or there has been a hard ICMP error [RFC1122] or a TLS alert. The sender knows a response was received when it receives an ACK with a sequence number that indicates it is a response to one of the transmissions of this messages. For example, assuming an RTO of 500 ms, requests would be sent at times 0 ms, 500 ms, 1500 ms, 3500 ms, and 7500 ms. If all retransmissions for a message fail, then the sending node SHOULD close the connection routing the message.
To determine when a link may be failing without waiting for the final timeout, observe when no ACKs have been received for an entire RTO interval, and then wait for three retransmissions to occur beyond that point. If no ACKs have been received by the time the third retransmission occurs, it is RECOMMENDED that the link be removed from the routing table. The link MAY be restored to the routing table if ACKs resume before the connection is closed, as described above.
Once an ACK has been received for a message, the next message can be sent, but the peer SHOULD ensure that there is at least 10 ms between sending any two messages. The only time a value less than 10 ms can be used is when it is known that all nodes are on a network that can support retransmissions faster than 10 ms with no congestion issues.
This overlay link protocol consists of DTLS over UDP while implementing the Simple Reliability protocol. STUN Connectivity checks and keepalives are used.
This overlay link protocol consists of TLS over TCP with the framing header. Because ICE is not used, STUN connectivity checks are not used upon establishing the TCP connection, nor are they used for keepalives.
Because the TCP layer's application-level timeout is too slow to be useful for overlay routing, the Overlay Link implementation MUST use the framing header to measure the RTT of the connection and calculate an RTO as specified in Section 2 of [RFC2988]. The resulting RTO is not used for retransmissions, but as a timeout to indicate when the link SHOULD be removed from the routing table. It is RECOMMENDED that such a connection be retained for 30s to determine if the failure was transient before concluding the link has failed permanently.
When sending candidates for TLS/TCP with FH, No-ICE, a passive candidate MUST be provided.
This overlay link protocol consists of DTLS over UDP while implementing the Simple Reliability protocol. Because ICE is not used, no STUN connectivity checks or keepalives are used.
In order to allow transmission over datagram protocols such as DTLS, RELOAD messages may be fragmented.
Any node along the path can fragment the message but only the final destination reassembles the fragments. When a node takes a packet and fragments it, each fragment has a full copy of the Forwarding Header but the data after the Forwarding Header is broken up in appropriate sized chunks. The size of the payload chunks needs to take into account space to allow the via and destination lists to grow. Each fragment MUST contain a full copy of the via and destination list and MUST contain at least 256 bytes of the message body. If the via and destination list are so large that this is not possible, RELOAD fragmentation is not performed and IP-layer fragmentation is allowed to occur. When a message must be fragmented, it SHOULD be split into equal-sized fragments that are no larger than the PMTU of the next overlay link minus 32 bytes. This is to allow the via list to grow before further fragmentation is required.
Note that this fragmentation is not optimal for the end-to-end path - a message may be refragmented multiple times as it traverses the overlay but is only assembled at the final destination. This option has been chosen as it is far easier to implement than e2e PMTU discovery across an ever-changing overlay, and it effectively addresses the reliability issues of relying on IP-layer fragmentation. However, PING can be used to allow e2e PMTU to be implemented if desired.
Upon receipt of a fragmented message by the intended peer, the peer holds the fragments in a holding buffer until the entire message has been received. The message is then reassembled into a single message and processed. In order to mitigate denial of service attacks, receivers SHOULD time out incomplete fragments after maximum request lifetime (15 seconds). Note this time was derived from looking at the end to end retransmission time and saving fragments long enough for the full end to end retransmissions to take place. Ideally the receiver would have enough buffer space to deal with as many fragments as can arrive in the maximum request lifetime. However, if the receiver runs out of buffer space to reassemble the messages it MUST drop the message.
When a message is fragmented, the fragment offset value is stored in the lower 24 bits of the fragment field of the forwarding header. The offset is the number of bytes between the end of the forwarding header and the start of the data. The first fragment therefore has an offset of 0. The first and last bit indicators MUST be appropriately set. If the message is not fragmented, then both the first and last fragment bits are set to 1 and the offset is 0 resulting in a fragment value of 0xC0000000. Note that this means that the first fragment bit is always 1, so isn't actually that useful.
RELOAD provides a set of generic mechanisms for storing and retrieving data in the Overlay Instance. These mechanisms can be used for new applications simply by defining new code points and a small set of rules. No new protocol mechanisms are required.
The basic unit of stored data is a single StoredData structure:
struct {
uint32 length;
uint64 storage_time;
uint32 lifetime;
StoredDataValue value;
Signature signature;
} StoredData;
The contents of this structure are as follows:
Each Resource-ID specifies a single location in the Overlay Instance. However, each location may contain multiple StoredData values distinguished by Kind-ID. The definition of a kind describes both the data values which may be stored and the data model of the data. Some data models allow multiple values to be stored under the same Kind-ID. Section Section 6.2 describes the available data models. Thus, for instance, a given Resource-ID might contain a single-value element stored under Kind-ID X and an array containing multiple values stored under Kind-ID Y.
Each StoredData element is individually signed. However, the signature also must be self-contained and cover the Kind-ID and Resource-ID even though they are not present in the StoredData structure. The input to the signature algorithm is:
Where || indicates concatenation.
Where these values are:
Once the signature has been computed, the signature is represented using a signature element, as described in Section 5.3.4.
The protocol currently defines the following data models:
These are represented with the StoredDataValue structure. The actual dataModel is known from the kind being stored.
struct {
Boolean exists;
opaque value<0..2^32-1>;
} DataValue;
struct {
select (dataModel) {
case single_value:
DataValue single_value_entry;
case array:
ArrayEntry array_entry;
case dictionary:
DictionaryEntry dictionary_entry;
/* This structure may be extended */
};
} StoredDataValue;
We now discuss the properties of each data model in turn:
A single-value element is a simple sequence of bytes. There may be only one single-value element for each Resource-ID, Kind-ID pair.
A single value element is represented as a DataValue, which contains the following two elements:
An array is a set of opaque values addressed by an integer index. Arrays are zero based. Note that arrays can be sparse. For instance, a Store of "X" at index 2 in an empty array produces an array with the values [ NA, NA, "X"]. Future attempts to fetch elements at index 0 or 1 will return values with "exists" set to False.
A array element is represented as an ArrayEntry:
struct {
uint32 index;
DataValue value;
} ArrayEntry;
The contents of this structure are:
A dictionary is a set of opaque values indexed by an opaque key with one value for each key. A single dictionary entry is represented as follows:
A dictionary element is represented as a DictionaryEntry:
typedef opaque DictionaryKey<0..2^16-1>;
struct {
DictionaryKey key;
DataValue value;
} DictionaryEntry;
The contents of this structure are:
Every kind which is storable in an overlay MUST be associated with an access control policy. This policy defines whether a request from a given node to operate on a given value should succeed or fail. It is anticipated that only a small number of generic access control policies are required. To that end, this section describes a small set of such policies and Section 13.4 establishes a registry for new policies if required. Each policy has a short string identifier which is used to reference it in the configuration document.
In the USER-MATCH policy, a given value MUST be written (or overwritten) if and only if the request is signed with a key associated with a certificate whose user name hashes (using the hash function for the overlay) to the Resource-ID for the resource. Recall that the certificate may, depending on the overlay configuration, be self-signed.
In the NODE-MATCH policy, a given value MUST be written (or overwritten) if and only if the request is signed with a key associated with a certificate whose Node-ID hashes (using the hash function for the overlay) to the Resource-ID for the resource.
The USER-NODE-MATCH policy may only be used with dictionary types. In the USER-NODE-MATCH policy, a given value MUST be written (or overwritten) if and only if the request is signed with a key associated with a certificate whose user name hashes (using the hash function for the overlay) to the Resource-ID for the resource. In addition, the dictionary key MUST be equal to the Node-ID in the certificate.
In the NODE-MULTIPLE policy, a given value MUST be written (or overwritten) if and only if the request is signed with a key associated with a certificate containing a Node-ID such that H(Node-ID || i) is equal to the Resource-ID for some small integer value of i. When this policy is in use, the maximum value of i MUST be specified in the kind definition.
Note that as i is not carried on the wire, the verifier MUST iterate through potential i values up to the maximum value in order to determine whether a store is acceptable.
RELOAD provides several methods for storing and retrieving data:
These methods are each described in the following sections.
The Store method is used to store data in the overlay. The format of the Store request depends on the data model which is determined by the kind.
A StoreReq message is a sequence of StoreKindData values, each of which represents a sequence of stored values for a given kind. The same Kind-ID MUST NOT be used twice in a given store request. Each value is then processed in turn. These operations MUST be atomic. If any operation fails, the state MUST be rolled back to before the request was received.
The store request is defined by the StoreReq structure:
struct {
KindId kind;
uint64 generation_counter;
StoredData values<0..2^32-1>;
} StoreKindData;
struct {
ResourceId resource;
uint8 replica_number;
StoreKindData kind_data<0..2^32-1>;
} StoreReq;
A single Store request stores data of a number of kinds to a single resource location. The contents of the structure are:
If the replica number is zero, then the peer MUST check that it is responsible for the resource and, if not, reject the request. If the replica number is nonzero, then the peer MUST check that it expects to be a replica for the resource and that the request sender is consistent with being the responsible node (i.e., that the receiving peer does not know of a better node) and, if not, reject the request.
Each StoreKindData element represents the data to be stored for a single Kind-ID. The contents of the element are:
The peer MUST perform the following checks:
If all these checks succeed, the peer MUST attempt to store the data values. For non-replica stores, if the store succeeds and the data is changed, then the peer must increase the generation counter by at least one. If there are multiple stored values in a single StoreKindData, it is permissible for the peer to increase the generation counter by only 1 for the entire Kind-ID, or by 1 or more than one for each value. Accordingly, all stored data values must have a generation counter of 1 or greater. 0 is used in the Store request to indicate that the generation counter should be ignored for processing this request; however the responsible peer should increase the stored generation counter and should return the correct generation counter in the response.
When a peer stores data previously stored by another node (e.g., for replicas or topology shifts) it MUST adjust the lifetime value downward to reflect the amount of time the value was stored at the peer. The adjustment SHOULD be implemented by an algorithm equivalent to the following: at the time the peer initially receives the StoreReq it notes the local time T. When it then attempts to do a StoreReq to another node it should decrement the lifetime value by the difference between the current local time and T.
Unless otherwise specified by the usage, if a peer attempts to store data previously stored by another node (e.g., for replicas or topology shifts) and that store fails with either an Error_Generation_Counter_Too_Low or an Error_Data_Too old error, the peer MUST fetch the newer data from the peer generating the error and use that to replace its own copy. This rule allows resynchronization after partitions heal.
The properties of stores for each data model are as follows:
The following figure shows the relationship between these structures for an example store which stores the following values at resource "1234"
Store
resource=1234
replica_number = 0
/ \
/ \
StoreKindData StoreKindData
kind=X (Single-Value) kind=Y (Array)
generation_counter = 99 generation_counter = 107
| /\
| / \
StoredData / \
storage_time = xxxxxxx / \
lifetime = 86400 / \
signature = XXXX / \
| | |
| StoredData StoredData
| storage_time = storage_time =
| yyyyyyyy zzzzzzz
| lifetime = 86400 lifetime = 33200
| signature = YYYY signature = ZZZZ
| | |
StoredDataValue | |
value="abc" | |
| |
StoredDataValue StoredDataValue
index=0 index=1
value="foo" value="bar"
In response to a successful Store request the peer MUST return a StoreAns message containing a series of StoreKindResponse elements containing the current value of the generation counter for each Kind-ID, as well as a list of the peers where the data will be replicated by the node processing the request.
struct {
KindId kind;
uint64 generation_counter;
NodeId replicas<0..2^16-1>;
} StoreKindResponse;
struct {
StoreKindResponse kind_responses<0..2^16-1>;
} StoreAns;
The contents of each StoreKindResponse are:
The response itself is just StoreKindResponse values packed end-to-end.
If any of the generation counters in the request precede the corresponding stored generation counter, then the peer MUST fail the entire request and respond with an Error_Generation_Counter_Too_Low error. The error_info in the ErrorResponse MUST be a StoreAns response containing the correct generation counter for each kind and the replica list, which will be empty. For original (non-replica) stores, a node which receives such an error SHOULD attempt to fetch the data and, if the storage_time value is newer, replace its own data with that newer data. This rule improves data consistency in the case of partitions and merges.
If the data being stored is too large for the allowed limit by the given usage, then the peer MUST fail the request and generate an Error_Data_Too_Large error.
If any type of request tries to access a data kind that the node does not know about, an Error_Unknown_Kind MUST be generated. The error_info in the Error_Response is:
KindId unknown_kinds<0..2^8-1>;
which lists all the kinds that were unrecognized. A node which receives this error MUST generate a ConfigUpdate message which contains the appropriate kind definition (assuming that in fact a kind was used which was defined in the configuration document).
This version of RELOAD (unlike previous versions) does not have an explicit Remove operation. Rather, values are Removed by storing "nonexistent" values in their place. Each DataValue contains a boolean value called "exists" which indicates whether a value is present at that location. In order to effectively remove a value, the owner stores a new DataValue with:
Storing nodes MUST treat these nonexistent values the same way they treat any other stored value, including overwriting the existing value, replicating them, and aging them out as necessary when lifetime expires. When a stored nonexistent value's lifetime expires, it is simply removed from the storing node like any other stored value expiration. Note that in the case of arrays and dictionaries, this may create an implicit, unsigned "nonexistent" value to represent a gap in the data structure. However, this value isn't persistent nor is it replicated. It is simply synthesized by the storing node.
The Fetch request retrieves one or more data elements stored at a given Resource-ID. A single Fetch request can retrieve multiple different kinds.
struct {
int32 first;
int32 last;
} ArrayRange;
struct {
KindId kind;
uint64 generation;
uint16 length;
select (dataModel) {
case single_value: ; /* Empty */
case array:
ArrayRange indices<0..2^16-1>;
case dictionary:
DictionaryKey keys<0..2^16-1>;
/* This structure may be extended */
} model_specifier;
} StoredDataSpecifier;
struct {
ResourceId resource;
StoredDataSpecifier specifiers<0..2^16-1>;
} FetchReq;
The contents of the Fetch requests are as follows:
Each StoredDataSpecifier specifies a single kind of data to retrieve and (if appropriate) the subset of values that are to be retrieved. The contents of the StoredDataSpecifier structure are as follows:
The model_specifier is as follows:
The generation counter is used to indicate the requester's expected state of the storing peer. If the generation counter in the request matches the stored counter, then the storing peer returns a response with no StoredData values.
Note that because the certificate for a user is typically stored at the same location as any data stored for that user, a requesting node that does not already have the user's certificate should request the certificate in the Fetch as an optimization.
The response to a successful Fetch request is a FetchAns message containing the data requested by the requester.
struct {
KindId kind;
uint64 generation;
StoredData values<0..2^32-1>;
} FetchKindResponse;
struct {
FetchKindResponse kind_responses<0..2^32-1>;
} FetchAns;
The FetchAns structure contains a series of FetchKindResponse structures. There MUST be one FetchKindResponse element for each Kind-ID in the request.
The contents of the FetchKindResponse structure are as follows:
There is one subtle point about signature computation on arrays. If the storing node uses the append feature (where the index=0xffffffff), then the index in the StoredData that is returned will not match that used by the storing node, which would break the signature. In order to avoid this issue, the index value in the array is set to zero before the signature is computed. This implies that malicious storing nodes can reorder array entries without being detected.
The Stat request is used to get metadata (length, generation counter, digest, etc.) for a stored element without retrieving the element itself. The name is from the UNIX stat(2) system call which performs a similar function for files in a file system. It also allows the requesting node to get a list of matching elements without requesting the entire element.
The Stat request is identical to the Fetch request. It simply specifies the elements to get metadata about.
struct {
ResourceId resource;
StoredDataSpecifier specifiers<0..2^16-1>;
} StatReq;
The Stat response contains the same sort of entries that a Fetch response would contain; however, instead of containing the element data it contains metadata.
struct {
Boolean exists;
uint32 value_length;
HashAlgorithm hash_algorithm;
opaque hash_value<0..255>;
} MetaData;
struct {
uint32 index;
MetaData value;
} ArrayEntryMeta;
struct {
DictionaryKey key;
MetaData value;
} DictionaryEntryMeta;
struct {
select (model) {
case single_value:
MetaData single_value_entry;
case array:
ArrayEntryMeta array_entry;
case dictionary:
DictionaryEntryMeta dictionary_entry;
/* This structure may be extended */
};
} MetaDataValue;
struct {
uint32 value_length;
uint64 storage_time;
uint32 lifetime;
MetaDataValue metadata;
} StoredMetaData;
struct {
KindId kind;
uint64 generation;
StoredMetaData values<0..2^32-1>;
} StatKindResponse;
struct {
StatKindResponse kind_responses<0..2^32-1>;
} StatAns;
The structures used in StatAns parallel those used in FetchAns: a response consists of multiple StatKindResponse values, one for each kind that was in the request. The contents of the StatKindResponse are the same as those in the FetchKindResponse, except that the values list contains StoredMetaData entries instead of StoredData entries.
The contents of the StoredMetaData structure are the same as the corresponding fields in StoredData except that there is no signature field and the value is a MetaDataValue rather than a StoredDataValue.
A MetaDataValue is a variant structure, like a StoredDataValue, except for the types of each arm, which replace DataValue with MetaData.
The only really new structure is MetaData, which has the following contents:
The Find request can be used to explore the Overlay Instance. A Find request for a Resource-ID R and a Kind-ID T retrieves the Resource-ID (if any) of the resource of kind T known to the target peer which is closest to R. This method can be used to walk the Overlay Instance by iteratively fetching R_n+1=nearest(1 + R_n).
The FindReq message contains a Resource-ID and a series of Kind-IDs identifying the resource the peer is interested in.
struct {
ResourceId resource;
KindId kinds<0..2^8-1>;
} FindReq;
The request contains a list of Kind-IDs which the Find is for, as indicated below:
A response to a successful Find request is a FindAns message containing the closest Resource-ID on the peer for each kind specified in the request.
struct {
KindId kind;
ResourceId closest;
} FindKindData;
struct {
FindKindData results<0..2^16-1>;
} FindAns;
If the processing peer is not responsible for the specified Resource-ID, it SHOULD return an Error_Not_Found error code.
For each Kind-ID in the request the response MUST contain a FindKindData indicating the closest Resource-ID for that Kind-ID, unless the kind is not allowed to be used with Find in which case a FindKindData for that Kind-ID MUST NOT be included in the response. If a Kind-ID is not known, then the corresponding Resource-ID MUST be 0. Note that different Kind-IDs may have different closest Resource-IDs.
The response is simply a series of FindKindData elements, one per kind, concatenated end-to-end. The contents of each element are:
Note that the response does not contain the contents of the data stored at these Resource-IDs. If the requester wants this, it must retrieve it using Fetch.
There are two ways to define a new kind. The first is by writing a document and registering the kind-id with IANA. This is the preferred method for kinds which may be widely used and reused. The second method is to simply define the kind and its parameters in the configuration document using the section of kind-id space set aside for private use. This method MAY be used to define ad hoc kinds in new overlays.
However a kind is defined, the definition must include:
In addition, when kinds are registered with IANA, each kind is assigned a short string name which is used to refer to it in configuration documents.
While each kind needs to define what data model is used for its data, that does not mean that it must define new data models. Where practical, kinds should use the existing data models. The intention is that the basic data model set be sufficient for most applications/usages.
The Certificate Store usage allows a peer to store its certificate in the overlay, thus avoiding the need to send a certificate in each message - a reference may be sent instead.
A user/peer MUST store its certificate at Resource-IDs derived from two Resource Names:
Note that in the second case the certificate is not stored at the peer's Node-ID but rather at a hash of the peer's Node-ID. The intention here (as is common throughout RELOAD) is to avoid making a peer responsible for its own data.
A peer MUST ensure that the user's certificates are stored in the Overlay Instance. New certificates are stored at the end of the list. This structure allows users to store an old and a new certificate that both have the same Node-ID, which allows for migration of certificates when they are renewed.
This usage defines the following kinds:
The TURN server usage allows a RELOAD peer to advertise that it is prepared to be a TURN server as defined in [RFC5766]. When a node starts up, it joins the overlay network and forms several connections in the process. If the ICE stage in any of these connections returns a reflexive address that is not the same as the peer's perceived address, then the peer is behind a NAT and not a candidate for a TURN server. Additionally, if the peer's IP address is in the private address space range, then it is also not a candidate for a TURN server. Otherwise, the peer SHOULD assume it is a potential TURN server and follow the procedures below.
If the node is a candidate for a TURN server it will insert some pointers in the overlay so that other peers can find it. The overlay configuration file specifies a turn-density parameter that indicates how many times each TURN server should record itself in the overlay. Typically this should be set to the reciprocal of the estimate of what percentage of peers will act as TURN servers. If the turn-density is not set to zero, for each value, called d, between 1 and turn-density, the peer forms a Resource Name by concatenating its Node-ID and the value d. This Resource Name is hashed to form a Resource-ID. The address of the peer is stored at that Resource-ID using type TURN-SERVICE and the TurnServer object:
struct {
uint8 iteration;
IpAddressAndPort server_address;
} TurnServer;
The contents of this structure are as follows:
Peers that provide this service need to support the TURN extensions to STUN for media relay as defined in [RFC5766].
This usage defines the following kind to indicate that a peer is willing to act as a TURN server:
Peers can find other servers by selecting a random Resource-ID and then doing a Find request for the appropriate Kind-ID with that Resource-ID. The Find request gets routed to a random peer based on the Resource-ID. If that peer knows of any servers, they will be returned. The returned response may be empty if the peer does not know of any servers, in which case the process gets repeated with some other random Resource-ID. As long as the ratio of servers relative to peers is not too low, this approach will result in finding a server relatively quickly.
This algorithm is assigned the name chord-reload to indicate it is an adaptation of the basic Chord DHT algorithm.
This algorithm differs from the originally presented Chord algorithm [Chord]. It has been updated based on more recent research results and implementation experiences, and to adapt it to the RELOAD protocol. A short list of differences:
The algorithm described here is a modified version of the Chord algorithm. Each peer keeps track of a finger table and a neighbor table. The neighbor table contains at least the three peers before and after this peer in the DHT ring. There may not be three entries in all cases such as small rings or while the ring topology is changing. The first entry in the finger table contains the peer half-way around the ring from this peer; the second entry contains the peer that is 1/4 of the way around; the third entry contains the peer that is 1/8th of the way around, and so on. Fundamentally, the chord data structure can be thought of a doubly-linked list formed by knowing the successors and predecessor peers in the neighbor table, sorted by the Node-ID. As long as the successor peers are correct, the DHT will return the correct result. The pointers to the prior peers are kept to enable the insertion of new peers into the list structure. Keeping multiple predecessor and successor pointers makes it possible to maintain the integrity of the data structure even when consecutive peers simultaneously fail. The finger table forms a skip list, so that entries in the linked list can be found in O(log(N)) time instead of the typical O(N) time that a linked list would provide.
A peer, n, is responsible for a particular Resource-ID k if k is less than or equal to n and k is greater than p, where p is the Node-ID of the previous peer in the neighbor table. Care must be taken when computing to note that all math is modulo 2^128.
For this Chord topology plugin, the size of the Resource-ID is 128 bits. The hash of a Resource-ID is computed using SHA-1 [RFC3174]then truncating the SHA-1 result to the most significant 128 bits.
The routing table is the union of the neighbor table and the finger table.
If a peer is not responsible for a Resource-ID k, but is directly connected to a node with Node-ID k, then it routes the message to that node. Otherwise, it routes the request to the peer in the routing table that has the largest Node-ID that is in the interval between the peer and k. If no such node is found, it finds the smallest Node-Id that is greater than k and routes the message to that node.
When a peer receives a Store request for Resource-ID k, and it is responsible for Resource-ID k, it stores the data and returns a success response. It then sends a Store request to its successor in the neighbor table and to that peer's successor. Note that these Store requests are addressed to those specific peers, even though the Resource-ID they are being asked to store is outside the range that they are responsible for. The peers receiving these check they came from an appropriate predecessor in their neighbor table and that they are in a range that this predecessor is responsible for, and then they store the data. They do not themselves perform further Stores because they can determine that they are not responsible for the Resource-ID.
Managing replicas as the overlay changes is described in Section 9.7.3.
The sequential replicas used in this overlay algorithm protect against peer failure but not against malicious peers. Additional replication from the Usage is required to protect resources from such attacks, as discussed in Section 12.5.4.
The join process for a joining party (JP) with Node-ID n is as follows.
If JP sends an Attach to AP with send_update, it immediately knows most of its expected neighbors from AP's routing table update and can directly connect to them. This is the RECOMMENDED procedure.
If for some reason JP does not get AP's routing table, it can still populate its neighbor table incrementally. It sends a Ping directed at Resource-ID n+1 (directly after its own Resource-ID). This allows it to discover its own successor. Call that node p0. It then sends a ping to p0+1 to discover its successor (p1). This process can be repeated to discover as many successors as desired. The values for the two peers before p will be found at a later stage when n receives an Update. An alternate procedure is to send Attaches to those nodes rather than pings, which forms the connections immediately but may be slower if the nodes need to collect ICE candidates, thus reducing parallelism.
In order to set up its finger table entry for peer i, JP simply sends an Attach to peer (n+2^(128-i). This will be routed to a peer in approximately the right location around the ring.
The joining peer MUST NOT send any Update message placing itself in the overlay until it has successfully completed an Attach with each peer that should be in its neighbor table.
When a peer needs to Attach to a new peer in its neighbor table, it MUST source-route the Attach request through the peer from which it learned the new peer's Node-ID. Source-routing these requests allows the overlay to recover from instability.
All other Attach requests, such as those for new finger table entries, are routed conventionally through the overlay.
A chord Update is defined as
enum { reserved (0),
peer_ready(1), neighbors(2), full(3), (255) }
ChordUpdateType;
struct {
uint32 uptime;
ChordUpdateType type;
select(type){
case peer_ready: /* Empty */
;
case neighbors:
NodeId predecessors<0..2^16-1>;
NodeId successors<0..2^16-1>;
case full:
NodeId predecessors<0..2^16-1>;
NodeId successors<0..2^16-1>;
NodeId fingers<0..2^16-1>;
};
} ChordUpdate;
The "uptime" field contains the time this peer has been up in seconds.
The "type" field contains the type of the update, which depends on the reason the update was sent.
If the message is of type "neighbors", then the contents of the message will be:
If the message is of type "full", then the contents of the message will be:
A peer MUST maintain an association (via Attach) to every member of its neighbor set. A peer MUST attempt to maintain at least three predecessors and three successors, even though this will not be possible if the ring is very small. It is RECOMMENDED that O(log(N)) predecessors and successors be maintained in the neighbor set.
Every time a connection to a peer in the neighbor table is lost (as determined by connectivity pings or the failure of some request), the peer MUST remove the entry from its neighbor table and replace it with the best match it has from the other peers in its routing table. If using reactive recovery, it then sends an immediate Update to all nodes in its Neighbor Table. The update will contain all the Node-IDs of the current entries of the table (after the failed one has been removed). Note that when replacing a successor the peer SHOULD delay the creation of new replicas for successor replacement hold-down time (30 seconds) after removing the failed entry from its neighbor table in order to allow a triggered update to inform it of a better match for its neighbor table.
If the neighbor failure effects the peer's range of responsible IDs, then the Update MUST be sent to all nodes in its Connection Table.
A peer MAY attempt to reestablish connectivity with a lost neighbor either by waiting additional time to see if connectivity returns or by actively routing a new Attach to the lost peer. Details for these procedures are beyond the scope of this document. In no event does an attempt to reestablish connectivity with a lost neighbor allow the peer to remain in the neighbor table. Such a peer is returned to the neighbor table once connectivity is reestablished.
If connectivity is lost to all successor peers in the neighbor table, then this peer should behave as if it is joining the network and use Pings to find a peer and send it a Join. If connectivity is lost to all the peers in the finger table, this peer should assume that it has been disconnected from the rest of the network, and it should periodically try to join the DHT.
If a finger table entry is found to have failed, all references to the failed peer are removed from the finger table and replaced with the closest preceding peer from the finger table or neighbor table.
If using reactive recovery, the peer initiates a search for a new finger table entry as described below.
When a peer, N, receives an Update request, it examines the Node-IDs in the UpdateReq and at its neighbor table and decides if this UpdateReq would change its neighbor table. This is done by taking the set of peers currently in the neighbor table and comparing them to the peers in the update request. There are two major cases:
In the first case, no change is needed.
In the second case, N MUST attempt to Attach to the new peers and if it is successful it MUST adjust its neighbor set accordingly. Note that it can maintain the now inferior peers as neighbors, but it MUST remember the closer ones.
After any Pings and Attaches are done, if the neighbor table changes and the peer is using reactive recovery, the peer sends an Update request to each member of its Connection Table. These Update requests are what end up filling in the predecessor/successor tables of peers that this peer is a neighbor to. A peer MUST NOT enter itself in its successor or predecessor table and instead should leave the entries empty.
If peer N is responsible for a Resource-ID R, and N discovers that the replica set for R (the next two nodes in its successor set) has changed, it MUST send a Store for any data associated with R to any new node in the replica set. It SHOULD NOT delete data from peers which have left the replica set.
When a peer N detects that it is no longer in the replica set for a resource R (i.e., there are three predecessors between N and R), it SHOULD delete all data associated with R from its local store.
When a peer discovers that its range of responsible IDs have changed, it MUST send an Update to all entries in its connection table.
There are four components to stabilization:
A peer MUST periodically send an Update request to every peer in its Connection Table. The purpose of this is to keep the predecessor and successor lists up to date and to detect failed peers. The default time is about every ten minutes, but the configuration server SHOULD set this in the configuration document using the "chord-update-interval" element (denominated in seconds.) A peer SHOULD randomly offset these Update requests so they do not occur all at once.
A peer MUST periodically search for new peers to replace invalid (repeated) entries in the finger table. A finger table entry i is valid if it is in the range [n+2^(128-i), n+2^(128-(i-1))-2^(128-(i+1))]. Invalid entries occur in the finger table when a previous finger table entry has failed or when no peer has been found in that range.
A peer SHOULD NOT send Ping requests looking for new finger table entries more often than the configuration element "chord-ping-interval", which defaults to 3600 seconds (one per hour).
Two possible methods for searching for new peers for the finger table entries are presented:
Alternative 1: A peer selects one entry in the finger table from among the invalid entries. It pings for a new peer for that finger table entry. The selection SHOULD be exponentially weighted to attempt to replace earlier (lower i) entries in the finger table. A simple way to implement this selection is to search through the finger table entries from i=0 and each time an invalid entry is encountered, send a Ping to replace that entry with probability 0.5.
Alternative 2: A peer monitors the Update messages received from its connections to observe when an Update indicates a peer that would be used to replace in invalid finger table entry, i, and flags that entry in the finger table. Every "chord-ping-interval" seconds, the peer selects from among those flagged candidates using an exponentially weighted probability as above.
When searching for a better entry, the peer SHOULD send the Ping to a Node-ID selected randomly from that range. Random selection is preferred over a search for strictly spaced entries to minimize the effect of churn on overlay routing [minimizing-churn-sigcomm06]. An implementation or subsequent specification MAY choose a method for selecting finger table entries other than choosing randomly within the range. Any such alternate methods SHOULD be employed only on finger table stabilization and not for the selection of initial finger table entries unless the alternative method is faster and imposes less overhead on the overlay.
A peer MAY choose to keep connections to multiple peers that can act for a given finger table entry.
If the finger table has less than 16 entries, the node SHOULD attempt to discover more fingers to grow the size of the table to 16. The value 16 was chosen to ensure high odds of a node maintaining connectivity to the overlay even with strange network partitions.
For many overlays, 16 finger table entries will be enough, but as an overlay grows very large, more than 16 entries may be required in the finger table for efficient routing. An implementation SHOULD be capable of increasing the number of entries in the finger table to 128 entries.
Note to implementers: Although log(N) entries are all that are required for optimal performance, careful implementation of stabilization will result in no additional traffic being generated when maintaining a finger table larger than log(N) entries. Implementers are encouraged to make use of RouteQuery and algorithms for determining where new finger table entries may be found. Complete details of possible implementations are outside the scope of this specification.
A simple approach to sizing the finger table is to ensure the finger table is large enough to contain at least the final successor in the peer's neighbor table.
To detect that a partitioning has occurred and to heal the overlay, a peer P MUST periodically repeat the discovery process used in the initial join for the overlay to locate an appropriate bootstrap node, B. P should then send a Ping for its own Node-ID routed through B. If a response is received from a peer S', which is not P's successor, then the overlay is partitioned and P should send an Attach to S' routed through B, followed by an Update sent to S'. (Note that S' may not be in P's neighbor table once the overlay is healed, but the connection will allow S' to discover appropriate neighbor entries for itself via its own stabilization.)
Future specifications may describe alternative mechanisms for determining when to repeat the discovery process.
For this topology plugin, the RouteQueryReq contains no additional information. The RouteQueryAns contains the single node ID of the next peer to which the responding peer would have routed the request message in recursive routing:
struct {
NodeId next_peer;
} ChordRouteQueryAns;
The contents of this structure are as follows:
If the requester has set the send_update flag, the responder SHOULD initiate an Update immediately after sending the RouteQueryAns.
enum { reserved (0),
from_succ(1), from_pred(2), (255) }
ChordLeaveType;
struct {
ChordLeaveType type;
select(type) {
case from_succ:
NodeId successors<0..2^16-1>;
case from_pred:
NodeId predecessors<0..2^16-1>;
};
} ChordLeaveData;
The 'type' field indicates whether the Leave request was sent by a
predecessor or a successor of the recipient:
from_succ
The Leave request was sent by a successor.
from_pred
The Leave request was sent by a predecessor.
If the type of the request is 'from_succ', the contents will be:
successors
The sender's successor list.
If the type of the request is 'from_pred', the contents will be:
predecessors
The sender's predecessor list.
To support extensions, such as [I-D.maenpaa-p2psip-self-tuning], Peers SHOULD send a Leave request to all members of their neighbor table prior to exiting the Overlay Instance. The overlay_specific_data field MUST contain the ChordLeaveData structure defined below:
Any peer which receives a Leave for a peer n in its neighbor set follows procedures as if it had detected a peer failure as described in Section 9.7.1.
The section defines the format of the configuration data as well the process to join a new overlay.
This specification defines a new content type "application/p2p-overlay+xml" for an MIME entity that contains overlay information. An example document is shown below.
<?xml version="1.0" encoding="UTF-8"?>
<overlay xmlns="urn:ietf:params:xml:ns:p2p:config-base"
xmlns:ext="urn:ietf:params:xml:ns:p2p:config-ext1"
xmlns:chord="urn:ietf:params:xml:ns:p2p:config-chord">
<configuration instance-name="overlay.example.org" sequence="22"
expiration="2002-10-10T07:00:00Z" ext:ext-example="stuff" >
<topology-plugin> CHORD-RELOAD </topology-plugin>
<node-id-length>16</node-id-length>
<root-cert>
MIIDJDCCAo2gAwIBAgIBADANBgkqhkiG9w0BAQUFADBwMQswCQYDVQQGEwJVUzET
MBEGA1UECBMKQ2FsaWZvcm5pYTERMA8GA1UEBxMIU2FuIEpvc2UxDjAMBgNVBAoT
BXNpcGl0MSkwJwYDVQQLEyBTaXBpdCBUZXN0IENlcnRpZmljYXRlIEF1dGhvcml0
eTAeFw0wMzA3MTgxMjIxNTJaFw0xMzA3MTUxMjIxNTJaMHAxCzAJBgNVBAYTAlVT
MRMwEQYDVQQIEwpDYWxpZm9ybmlhMREwDwYDVQQHEwhTYW4gSm9zZTEOMAwGA1UE
ChMFc2lwaXQxKTAnBgNVBAsTIFNpcGl0IFRlc3QgQ2VydGlmaWNhdGUgQXV0aG9y
aXR5MIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQDDIh6DkcUDLDyK9BEUxkud
+nJ4xrCVGKfgjHm6XaSuHiEtnfELHM+9WymzkBNzZpJu30yzsxwfKoIKugdNUrD4
N3viCicwcN35LgP/KnbN34cavXHr4ZlqxH+OdKB3hQTpQa38A7YXdaoz6goW2ft5
Mi74z03GNKP/G9BoKOGd5QIDAQABo4HNMIHKMB0GA1UdDgQWBBRrRhcU6pR2JYBU
bhNU2qHjVBShtjCBmgYDVR0jBIGSMIGPgBRrRhcU6pR2JYBUbhNU2qHjVBShtqF0
pHIwcDELMAkGA1UEBhMCVVMxEzARBgNVBAgTCkNhbGlmb3JuaWExETAPBgNVBAcT
CFNhbiBKb3NlMQ4wDAYDVQQKEwVzaXBpdDEpMCcGA1UECxMgU2lwaXQgVGVzdCBD
ZXJ0aWZpY2F0ZSBBdXRob3JpdHmCAQAwDAYDVR0TBAUwAwEB/zANBgkqhkiG9w0B
AQUFAAOBgQCWbRvv1ZGTRXxbH8/EqkdSCzSoUPrs+rQqR0xdQac9wNY/nlZbkR3O
qAezG6Sfmklvf+DOg5RxQq/+Y6I03LRepc7KeVDpaplMFGnpfKsibETMipwzayNQ
QgUf4cKBiF+65Ue7hZuDJa2EMv8qW4twEhGDYclpFU9YozyS1OhvUg==
</root-cert>
<root-cert> YmFkIGNlcnQK </root-cert>
<enrollment-server>https://example.org</enrollment-server>
<enrollment-server>https://example.net</enrollment-server>
<self-signed-permitted
digest="sha1">false</self-signed-permitted>
<bootstrap-node address="192.0.0.1" port="6084" />
<bootstrap-node address="192.0.2.2" port="6084" />
<bootstrap-node address="2001:DB8::1" port="6084" />
<turn-density> 20 </turn-density>
<multicast-bootstrap address="192.0.0.1" />
<multicast-bootstrap address="233.252.0.1" port="6084" />
<clients-permitted> false </clients-permitted>
<no-ice> false </no-ice>
<chord:chord-update-interval>
400</chord:chord-update-interval>
<chord:chord-ping-interval>30</chord:chord-ping-interval>
<chord:chord-reactive> true </chord:chord-reactive>
<shared-secret> password </shared-secret>
<max-message-size>4000</max-message-size>
<initial-ttl> 30 </initial-ttl>
<overlay-link-protocol>TLS</overlay-link-protocol>
<kind-signer> 47112162e84c69ba </kind-signer>
<kind-signer> 6eba45d31a900c06 </kind-signer>
<bad-node> 6ebc45d31a900c06 </bad-node>
<bad-node> 6ebc45d31a900ca6 </bad-node>
<ext:example-extension> foo </ext:example-extension>
<required-kinds>
<kind-block>
<kind name="SIP-REGISTRATION">
<data-model>SINGLE</data-model>
<access-control>USER-MATCH</access-control>
<max-count>1</max-count>
<max-size>100</max-size>
</kind>
<kind-signature>
VGhpcyBpcyBub3QgcmlnaHQhCg==
</kind-signature>
</kind-block>
<kind-block>
<kind id="2000">
<data-model>ARRAY</data-model>
<access-control>NODE-MULTIPLE</access-control>
<max-node-multiple>3</max-node-multiple>
<max-count>22</max-count>
<max-size>4</max-size>
<ext:example-kind-extension> 1
</ext:example-kind-extension>
</kind>
<kind-signature>
VGhpcyBpcyBub3QgcmlnaHQhCg==
</kind-signature>
</kind-block>
</required-kinds>
</configuration>
<signature> VGhpcyBpcyBub3QgcmlnaHQhCg== </signature>
<configuration instance-name="other.example.net">
</configuration>
<signature> VGhpcyBpcyBub3QgcmlnaHQhCg== </signature>
</overlay>
The file MUST be a well formed XML document and it SHOULD contain an encoding declaration in the XML declaration. If the charset parameter of the MIME content type declaration is present and it is different from the encoding declaration, the charset parameter takes precedence. Every application conforming to this specification MUST accept the UTF-8 character encoding to ensure minimal interoperability. The namespace for the elements defined in this specification is urn:ietf:params:xml:ns:p2p:config-base and urn:ietf:params:xml:ns:p2p:config-chord".
The file can contain multiple "configuration" elements where each one contains the configuration information for a different overlay. Each "configuration" has the following attributes:
Inside each overlay element, the following elements can occur:
Inside each overlay element, the required-kinds elements can also occur. This element indicates the kinds that members must support and contains multiple kind-block elements that each define a single kind that MUST be supported by nodes in the overlay. Each kind-block consists of a single kind element and a kind-signature. The kind element defines the kind. The kind-signature is the signature computed over the kind element.
Each kind has either an id attribute or a name attribute. The name attribute is a string representing the kind (the name registered to IANA) while the id is an integer kind-id allocated out of private space.
In addition, the kind element contains the following elements:
All of the non optional values MUST be provided. If the kind is registered with IANA, the data-model and access-control elements MUST match those in the kind registration, and clients MUST ignore them in favor of the IANA versions. Multiple required-kinds elements MAY be present.
The kind-block element also MUST contain a "kind-signature" element. This signature is computed across the kind from the beginning of the first < of the kind to the end of the last > of the kind in the same way as the signature element described later in this section.
The configuration file is a binary file and cannot be changed - including whitespace changes - or the signature will break. The signature is computed by taking each configuration element and starting from, and including, the first < at the start of <configuration> up to and including the > in </configuration> and treating this as a binary blob that is signed using the standard SecurityBlock defined in Section 5.3.4. The SecurityBlock is base 64 encoded using the base64 alphabet from RFC[RFC4648] and put in the signature element following the configuration object in the configuration file.
When a node receives a new configuration file, it MUST change its configuration to meet the new requirements. This may require the node to exit the DHT and re-join. If a node is not capable of supporting the new requirements, it MUST exit the overlay. If some information about a particular kind changes from what the node previously knew about the kind (for example the max size), the new information in the configuration files overrides any previously learned information. If any kind data was signed by a node that is no longer allowed to sign kinds, that kind MUST be discarded along with any stored information of that kind. Note that forcing an avalanche restart of the overlay with a configuration change that requires re-joining the overlay may result in serious performance problems, including total collapse of the network if configuration parameters are not properly considered. Such an event may be necessary in case of a compromised CA or similar problem, but for large overlays should be avoided in almost all circumstances.
The grammar for the configuration data is:
namespace chord = "urn:ietf:params:xml:ns:p2p:config-chord"
namespace local = ""
default namespace p2pcf = "urn:ietf:params:xml:ns:p2p:config-base"
namespace rng = "http://relaxng.org/ns/structure/1.0"
anything =
(element * { anything }
| attribute * { text }
| text)*
foreign-elements = element * - (p2pcf:* | local:* | chord:*)
{ anything }*
foreign-attributes = attribute * - (p2pcf:*|local:*|chord:*)
{ text }*
foreign-nodes = (foreign-attributes | foreign-elements)*
start = element p2pcf:overlay {
overlay-element
}
overlay-element &= element configuration {
attribute instance-name { xsd:string },
attribute expiration { xsd:dateTime }?,
attribute sequence { xsd:long }?,
foreign-attributes*,
parameter
}+
overlay-element &= element signature {
attribute algorithm { signature-algorithm-type }?,
xsd:base64Binary
}*
signature-algorithm-type |= "rsa-sha1"
signature-algorithm-type |= xsd:string # signature alg extensions
parameter &= element topology-plugin { topology-plugin-type }?
topology-plugin-type |= xsd:string # topo plugin extensions
parameter &= element max-message-size { xsd:unsignedInt }?
parameter &= element initial-ttl { xsd:int }?
parameter &= element root-cert { xsd:base64Binary }*
parameter &= element required-kinds { kind-block* }?
parameter &= element enrollment-server { xsd:anyURI }*
parameter &= element kind-signer { xsd:string }*
parameter &= element bad-node { xsd:string }*
parameter &= element no-ice { xsd:boolean }?
parameter &= element shared-secret { xsd:string }?
parameter &= element overlay-link-protocol { xsd:string }*
parameter &= element clients-permitted { xsd:boolean }?
parameter &= element turn-density { xsd:int }?
parameter &= element node-id-length { xsd:int }?
parameter &= foreign-elements*
parameter &=
element self-signed-permitted {
attribute digest { self-signed-digest-type },
xsd:boolean
}?
self-signed-digest-type |= "sha1"
self-signed-digest-type |= xsd:string # signature digest extensions
parameter &= element bootstrap-node {
attribute address { xsd:string },
attribute port { xsd:int }?
}*
parameter &= element multicast-bootstrap {
attribute address { xsd:string },
attribute port { xsd:int }?
}*
kind-block = element kind-block {
element kind {
( attribute name { kind-names }
| attribute id { xsd:int } ),
kind-parameter
} &
element kind-signature {
attribute algorithm { signature-algorithm-type }?,
xsd:base64Binary
}?
}
kind-parameter &= element max-count { xsd:int }
kind-parameter &= element max-size { xsd:int }
kind-parameter &= element max-node-multiple { xsd:int }?
kind-parameter &= element data-model { data-model-type }
data-model-type |= "SINGLE"
data-model-type |= "ARRAY"
data-model-type |= "DICTIONARY"
data-model-type |= xsd:string # data model extensions
kind-parameter &= element access-control { access-control-type }
access-control-type |= "USER-MATCH"
access-control-type |= "NODE-MATCH"
access-control-type |= "USER-NODE-MATCH"
access-control-type |= "NODE-MULTIPLE"
access-control-type |= xsd:string # access control extensions
kind-parameter &= foreign-elements*
kind-names |= "TURN-SERVICE"
kind-names |= "CERTIFICATE_BY_NODE"
kind-names |= "CERTIFICATE_BY_USER"
kind-names |= xsd:string # kind extensions
# Chord specific parameters
topology-plugin-type |= "CHORD-RELOAD"
parameter &= element chord:chord-ping-interval { xsd:int }?
parameter &= element chord:chord-update-interval { xsd:int }?
parameter &= element chord:chord-reactive { xsd:boolean }?
When a node first enrolls in a new overlay, it starts with a discovery process to find a configuration server.
The node first determines the overlay name. This value is provided by the user or some other out of band provisioning mechanism. The out of band mechanisms may also provide an optional URL for the configuration server. If a URL for the configuration server is not provided, the node MUST do a DNS SRV query using a Service name of "p2psip-enroll" and a protocol of TCP to find a configuration server and form the URL by appending a path of "/.well-known/p2psip-enroll" to the overlay name. This uses the "well known URI" framework defined in [RFC5785]. For example, if the overlay name was example.com, the URL would be "https://example.com//.well-known/p2psip-enroll".
Once an address and URL for the configuration server is determined, the peer forms an HTTPS connection to that IP address. The certificate MUST match the overlay name as described in [RFC2818]. Then the node MUST fetch a new copy of the configuration file. To do this, the peer performs a GET to the URL. The result of the HTTP GET is an XML configuration file described above, which replaces any previously learned configuration file for this overlay.
For overlays that do not use a configuration server, nodes obtain the configuration information needed to join the overlay through some out of band approach such an XML configuration file sent over email.
If the configuration document contains a enrollment-server element, credentials are required to join the Overlay Instance. A peer which does not yet have credentials MUST contact the enrollment server to acquire them.
RELOAD defines its own trivial certificate request protocol. We would have liked to have used an existing protocol but were concerned about the implementation burden of even the simplest of those protocols, such as [RFC5272] and [RFC5273]. Our objective was to have a protocol which could be easily implemented in a Web server which the operator did not control (e.g., in a hosted service) and was compatible with the existing certificate handling tooling as used with the Web certificate infrastructure. This means accepting bare PKCS#10 requests and returning a single bare X.509 certificate. Although the MIME types for these objects are defined, none of the existing protocols support exactly this model.
The certificate request protocol is performed over HTTPS. The request is an HTTP POST with the following properties:
The enrollment server MUST authenticate the request using the provided user name and password. If the authentication succeeds and the requested user name is acceptable, the server generates and returns a certificate. The SubjectAltName field in the certificate contains the following values:
The certificate is returned as type "application/pkix-cert" as defined in [RFC2585], with an HTTP status code of 200 OK. Certificate processing errors should be treated as HTTP errors and have appropriate HTTP status codes.
The client MUST check that the certificate returned was signed by one of the certificates received in the "root-cert" list of the overlay configuration data. The node then reads the certificate to find the Node-IDs it can use.
If the "self-signed-permitted" element is present in the configuration and set to "true", then a node MUST generate its own self-signed certificate to join the overlay. The self-signed certificate MAY contain any user name of the users choice.
The Node-ID MUST be computed by applying the digest specified in the self-signed-permitted element to the DER representation of the user's public key (more specifically the subjectPublicKeyInfo) and taking the high order bits. When accepting a self-signed certificate, nodes MUST check that the Node-ID and public keys match. This prevents Node-ID theft.
Once the node has constructed a self-signed certificate, it MAY join the overlay. Before storing its certificate in the overlay (Section 7) it SHOULD look to see if the user name is already taken and if so choose another user name. Note that this only provides protection against accidental name collisions. Name theft is still possible. If protection against name theft is desired, then the enrollment service must be used.
If no cached bootstrap nodes are available and the configuration file has an multicast-bootstrap element, then the node SHOULD send a Ping request over UDP to the address and port found to each multicast-bootstrap element found in the configuration document. This MAY be a multicast, broadcast, or anycast address. The Ping should use the wildcard Node-ID as the destination Node-ID.
The responder node that receives the Ping request SHOULD check that the overlay name is correct and that the requester peer sending the request has appropriate credentials for the overlay before responding to the Ping request even if the response is only an error.
In order to join the overlay, the joining node MUST contact a node in the overlay. Typically this means contacting the bootstrap nodes, since they are reachable by the local peer or have public IP addresses. If the joining node has cached a list of peers it has previously been connected with in this overlay, as an optimization it MAY attempt to use one or more of them as bootstrap nodes before falling back to the bootstrap nodes listed in the configuration file.
When contacting a bootstrap node, the joining node first forms the DTLS or TLS connection to the bootstrap node and then sends an Attach request over this connection with the destination Node-ID set to the joining node's Node-ID.
When the requester node finally does receive a response from some responding node, it can note the Node-ID in the response and use this Node-ID to start sending requests to join the Overlay Instance as described in Section 5.4.
After a node has successfully joined the overlay network, it will have direct connections to several peers. Some MAY be added to the cached bootstrap nodes list and used in future boots. Peers that are not directly connected MUST NOT be cached. The suggested number of peers to cache is 10. Algorithms for determining which peers to cache are beyond the scope of this specification.
The following abbreviation are used in the message flow diagrams: JP = joining peer, AP = admitting peer, NP = next peer after the AP, NNP = next next peer which is the peer after NP, PP = previous peer before the AP, PPP = previous previous peer which is the peer before the PP, BP = bootstrap peer.
In the following example, we assume that JP has formed a connection to one of the bootstrap nodes. JP then sends an Attach through that peer to a resource ID of itself (JP). It gets routed to the admitting peer (AP) because JP is not yet part of the overlay. When AP responds, JP and AP use ICE to set up a connection and then set up TLS. Once AP has connected to JP, AP sends to JP an Update to populate its Routing Table. The following example shows the Update happening after the TLS connection is formed but it could also happen before in which case the Update would often be routed through other nodes.
JP PPP PP AP NP NNP BP
| | | | | | |
| | | | | | |
| | | | | | |
|Attach Dest=JP | | | | |
|---------------------------------------------------------->|
| | | | | | |
| | | | | | |
| | |Attach Dest=JP | | |
| | |<--------------------------------------|
| | | | | | |
| | | | | | |
| | |Attach Dest=JP | | |
| | |-------->| | | |
| | | | | | |
| | | | | | |
| | |AttachAns | | |
| | |<--------| | | |
| | | | | | |
| | | | | | |
| | |AttachAns | | |
| | |-------------------------------------->|
| | | | | | |
| | | | | | |
|AttachAns | | | | |
|<----------------------------------------------------------|
| | | | | | |
| | | | | | |
|TLS | | | | | |
|.............................| | | |
| | | | | | |
| | | | | | |
| | | | | | |
|Update | | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|UpdateAns| | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
| | | | | | |
The JP then forms connections to the appropriate neighbors, such as NP, by sending an Attach which gets routed via other nodes. When NP responds, JP and NP use ICE and TLS to set up a connection.
JP PPP PP AP NP NNP BP
| | | | | | |
| | | | | | |
| | | | | | |
|Attach NP | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
| | | |Attach NP| | |
| | | |-------->| | |
| | | | | | |
| | | | | | |
| | | |AttachAns| | |
| | | |<--------| | |
| | | | | | |
| | | | | | |
|AttachAns | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|Attach | | | | | |
|-------------------------------------->| | |
| | | | | | |
| | | | | | |
|TLS | | | | | |
|.......................................| | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
JP also needs to populate its finger table (for Chord). It issues an Attach to a variety of locations around the overlay. The diagram below shows it sending an Attach halfway around the Chord ring to the JP + 2^127.
JP NP XX TP
| | | |
| | | |
| | | |
|Attach JP+2<<126 | |
|-------->| | |
| | | |
| | | |
| |Attach JP+2<<126 |
| |-------->| |
| | | |
| | | |
| | |Attach JP+2<<126
| | |-------->|
| | | |
| | | |
| | |AttachAns|
| | |<--------|
| | | |
| | | |
| |AttachAns| |
| |<--------| |
| | | |
| | | |
|AttachAns| | |
|<--------| | |
| | | |
| | | |
|TLS | | |
|.............................|
| | | |
| | | |
| | | |
| | | |
Once JP has a reasonable set of connections, it is ready to take its place in the DHT. It does this by sending a Join to AP. AP does a series of Store requests to JP to store the data that JP will be responsible for. AP then sends JP an Update explicitly labeling JP as its predecessor. At this point, JP is part of the ring and responsible for a section of the overlay. AP can now forget any data which is assigned to JP and not AP.
JP PPP PP AP NP NNP BP
| | | | | | |
| | | | | | |
| | | | | | |
|JoinReq | | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
|JoinAns | | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|StoreReq Data A | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|StoreAns | | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
|StoreReq Data B | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|StoreAns | | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
|UpdateReq| | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|UpdateAns| | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
In Chord, JP's neighbor table needs to contain its own predecessors. It couldn't connect to them previously because it did not yet know their addresses. However, now that it has received an Update from AP, it has AP's predecessors, which are also its own, so it sends Attaches to them. Below it is shown connecting to AP's closest predecessor, PP.
JP PPP PP AP NP NNP BP
| | | | | | |
| | | | | | |
| | | | | | |
|Attach Dest=PP | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
| | |Attach Dest=PP | | |
| | |<--------| | | |
| | | | | | |
| | | | | | |
| | |AttachAns| | | |
| | |-------->| | | |
| | | | | | |
| | | | | | |
|AttachAns| | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|TLS | | | | | |
|...................| | | | |
| | | | | | |
| | | | | | |
|UpdateReq| | | | | |
|------------------>| | | | |
| | | | | | |
| | | | | | |
|UpdateAns| | | | | |
|<------------------| | | | |
| | | | | | |
| | | | | | |
|UpdateReq| | | | | |
|---------------------------->| | | |
| | | | | | |
| | | | | | |
|UpdateAns| | | | | |
|<----------------------------| | | |
| | | | | | |
| | | | | | |
|UpdateReq| | | | | |
|-------------------------------------->| | |
| | | | | | |
| | | | | | |
|UpdateAns| | | | | |
|<--------------------------------------| | |
| | | | | | |
| | | | | | |
Finally, now that JP has a copy of all the data and is ready to route messages and receive requests, it sends Updates to everyone in its Routing Table to tell them it is ready to go. Below, it is shown sending such an update to TP.
JP NP XX TP
| | | |
| | | |
| | | |
|Update | | |
|---------------------------->|
| | | |
| | | |
|UpdateAns| | |
|<----------------------------|
| | | |
| | | |
| | | |
| | | |
RELOAD provides a generic storage service, albeit one designed to be useful for P2PSIP. In this section we discuss security issues that are likely to be relevant to any usage of RELOAD. More background information can be found in [RFC5765].
In any Overlay Instance, any given user depends on a number of peers with which they have no well-defined relationship except that they are fellow members of the Overlay Instance. In practice, these other nodes may be friendly, lazy, curious, or outright malicious. No security system can provide complete protection in an environment where most nodes are malicious. The goal of security in RELOAD is to provide strong security guarantees of some properties even in the face of a large number of malicious nodes and to allow the overlay to function correctly in the face of a modest number of malicious nodes.
P2PSIP deployments require the ability to authenticate both peers and resources (users) without the active presence of a trusted entity in the system. We describe two mechanisms. The first mechanism is based on public key certificates and is suitable for general deployments. The second is an admission control mechanism based on an overlay-wide shared symmetric key.
The two basic functions provided by overlay nodes are storage and routing: some node is responsible for storing a peer's data and for allowing a third peer to fetch this stored data. Other nodes are responsible for routing messages to and from the storing nodes. Each of these issues is covered in the following sections.
P2P overlays are subject to attacks by subversive nodes that may attempt to disrupt routing, corrupt or remove user registrations, or eavesdrop on signaling. The certificate-based security algorithms we describe in this specification are intended to protect overlay routing and user registration information in RELOAD messages.
To protect the signaling from attackers pretending to be valid peers (or peers other than themselves), the first requirement is to ensure that all messages are received from authorized members of the overlay. For this reason, RELOAD transports all messages over a secure channel (TLS and DTLS are defined in this document) which provides message integrity and authentication of the directly communicating peer. In addition, messages and data are digitally signed with the sender's private key, providing end-to-end security for communications.
This specification stores users' registrations and possibly other data in an overlay network. This requires a solution to securing this data as well as securing, as well as possible, the routing in the overlay. Both types of security are based on requiring that every entity in the system (whether user or peer) authenticate cryptographically using an asymmetric key pair tied to a certificate.
When a user enrolls in the Overlay Instance, they request or are assigned a unique name, such as "alice@dht.example.net". These names are unique and are meant to be chosen and used by humans much like a SIP Address of Record (AOR) or an email address. The user is also assigned one or more Node-IDs by the central enrollment authority. Both the name and the Node-ID are placed in the certificate, along with the user's public key.
Each certificate enables an entity to act in two sorts of roles:
Note that since only users of this Overlay Instance need to validate a certificate, this usage does not require a global PKI. Instead, certificates are signed by a central enrollment authority which acts as the certificate authority for the Overlay Instance. This authority signs each peer's certificate. Because each peer possesses the CA's certificate (which they receive on enrollment) they can verify the certificates of the other entities in the overlay without further communication. Because the certificates contain the user/peer's public key, communications from the user/peer can be verified in turn.
If self-signed certificates are used, then the security provided is significantly decreased, since attackers can mount Sybil attacks. In addition, attackers cannot trust the user names in certificates (though they can trust the Node-IDs because they are cryptographically verifiable). This scheme may be appropriate for some small deployments, such as a small office or an ad hoc overlay set up among participants in a meeting where all hosts on the network are trusted. Some additional security can be provided by using the shared secret admission control scheme as well.
Because all stored data is signed by the owner of the data the storing peer can verify that the storer is authorized to perform a store at that Resource-ID and also allow any consumer of the data to verify the provenance and integrity of the data when it retrieves it.
Note that RELOAD does not itself provide a revocation/status mechanism (though certificates may of course include OCSP responder information). Thus, certificate lifetimes should be chosen to balance the compromise window versus the cost of certificate renewal. Because RELOAD is already designed to operate in the face of some fraction of malicious peers, this form of compromise is not fatal.
All implementations MUST implement certificate-based security.
RELOAD also supports a shared secret admission control scheme that relies on a single key that is shared among all members of the overlay. It is appropriate for small groups that wish to form a private network without complexity. In shared secret mode, all the peers share a single symmetric key which is used to key TLS-PSK [RFC4279] or TLS-SRP [RFC5054] mode. A peer which does not know the key cannot form TLS connections with any other peer and therefore cannot join the overlay.
One natural approach to a shared-secret scheme is to use a user-entered password as the key. The difficulty with this is that in TLS-PSK mode, such keys are very susceptible to dictionary attacks. If passwords are used as the source of shared-keys, then TLS-SRP is a superior choice because it is not subject to dictionary attacks.
When certificate-based security is used in RELOAD, any given Resource-ID/Kind-ID pair is bound to some small set of certificates. In order to write data, the writer must prove possession of the private key for one of those certificates. Moreover, all data is stored, signed with the same private key that was used to authorize the storage. This set of rules makes questions of authorization and data integrity - which have historically been thorny for overlays - relatively simple.
When a client wants to store some value, it first digitally signs the value with its own private key. It then sends a Store request that contains both the value and the signature towards the storing peer (which is defined by the Resource Name construction algorithm for that particular kind of value).
When the storing peer receives the request, it must determine whether the storing client is authorized to store at this Resource-ID/Kind-ID pair. Determining this requires comparing the user's identity to the requirements of the access control model (see Section 6.3). If it satisfies those requirements the user is authorized to write, pending quota checks as described in the next section.
For example, consider the certificate with the following properties:
User name: alice@dht.example.com
Node-ID: 013456789abcdef
Serial: 1234
If Alice wishes to Store a value of the "SIP Location" kind, the Resource Name will be the SIP AOR "sip:alice@dht.example.com". The Resource-ID will be determined by hashing the Resource Name. Because SIP Location uses the USER-NODE-MATCH policy, it first verifies that the user name in the certificate hashes to the requested Resource-ID. It then verifies that the Node-Id in the certificate matches the dictionary key being used for the store. If both of these checks succeed, the Store is authorized. Note that because the access control model is different for different kinds, the exact set of checks will vary.
Being a peer in an Overlay Instance carries with it the responsibility to store data for a given region of the Overlay Instance. However, allowing clients to store unlimited amounts of data would create unacceptable burdens on peers and would also enable trivial denial of service attacks. RELOAD addresses this issue by requiring configurations to define maximum sizes for each kind of stored data. Attempts to store values exceeding this size MUST be rejected (if peers are inconsistent about this, then strange artifacts will happen when the zone of responsibility shifts and a different peer becomes responsible for overlarge data). Because each Resource-ID/Kind-ID pair is bound to a small set of certificates, these size restrictions also create a distributed quota mechanism, with the quotas administered by the central configuration server.
Allowing different kinds of data to have different size restrictions allows new usages the flexibility to define limits that fit their needs without requiring all usages to have expansive limits.
Because each stored value is signed, it is trivial for any retrieving peer to verify the integrity of the stored value. Some more care needs to be taken to prevent version rollback attacks. Rollback attacks on storage are prevented by the use of store times and lifetime values in each store. A lifetime represents the latest time at which the data is valid and thus limits (though does not completely prevent) the ability of the storing node to perform a rollback attack on retrievers. In order to prevent a rollback attack at the time of the Store request, we require that storage times be monotonically increasing. Storing peers MUST reject Store requests with storage times smaller than or equal to those they are currently storing. In addition, a fetching node which receives a data value with a storage time older than the result of the previous fetch knows a rollback has occurred.
The mechanisms described here provides a high degree of security, but some attacks remain possible. Most simply, it is possible for storing nodes to refuse to store a value (i.e., reject any request). In addition, a storing node can deny knowledge of values which it has previously accepted. To some extent these attacks can be ameliorated by attempting to store to/retrieve from replicas, but a retrieving client does not know whether it should try this or not, since there is a cost to doing so.
The certificate-based authentication scheme prevents a single peer from being able to forge data owned by other peers. Furthermore, although a subversive peer can refuse to return data resources for which it is responsible, it cannot return forged data because it cannot provide authentication for such registrations. Therefore parallel searches for redundant registrations can mitigate most of the effects of a compromised peer. The ultimate reliability of such an overlay is a statistical question based on the replication factor and the percentage of compromised peers.
In addition, when a kind is multivalued (e.g., an array data model), the storing node can return only some subset of the values, thus biasing its responses. This can be countered by using single values rather than sets, but that makes coordination between multiple storing agents much more difficult. This is a trade off that must be made when designing any usage.
Because the storage security system guarantees (within limits) the integrity of the stored data, routing security focuses on stopping the attacker from performing a DOS attack that misroutes requests in the overlay. There are a few obvious observations to make about this. First, it is easy to ensure that an attacker is at least a valid peer in the Overlay Instance. Second, this is a DOS attack only. Third, if a large percentage of the peers on the Overlay Instance are controlled by the attacker, it is probably impossible to perfectly secure against this.
In general, attacks on DHT routing are mounted by the attacker arranging to route traffic through one or two nodes it controls. In the Eclipse attack [Eclipse] the attacker tampers with messages to and from nodes for which it is on-path with respect to a given victim node. This allows it to pretend to be all the nodes that are reachable through it. In the Sybil attack [Sybil], the attacker registers a large number of nodes and is therefore able to capture a large amount of the traffic through the DHT.
Both the Eclipse and Sybil attacks require the attacker to be able to exercise control over her Node-IDs. The Sybil attack requires the creation of a large number of peers. The Eclipse attack requires that the attacker be able to impersonate specific peers. In both cases, these attacks are limited by the use of centralized, certificate-based admission control.
Admission to a RELOAD Overlay Instance is controlled by requiring that each peer have a certificate containing its Node-Id. The requirement to have a certificate is enforced by using certificate-based mutual authentication on each connection. (Note: the following only applies when self-signed certificates are not used.) Whenever a peer connects to another peer, each side automatically checks that the other has a suitable certificate. These Node-Ids are randomly assigned by the central enrollment server. This has two benefits:
The first property allows protection against Sybil attacks (provided the enrollment server uses strict rate limiting policies). The second property deters but does not completely prevent Eclipse attacks. Because an Eclipse attacker must impersonate peers on the other side of the attacker, he must have a certificate for suitable Node-Ids, which requires him to repeatedly query the enrollment server for new certificates, which will match only by chance. From the attacker's perspective, the difficulty is that if he only has a small number of certificates, the region of the Overlay Instance he is impersonating appears to be very sparsely populated by comparison to the victim's local region.
In general, whenever a peer engages in overlay activity that might affect the routing table it must establish its identity. This happens in two ways. First, whenever a peer establishes a direct connection to another peer it authenticates via certificate-based mutual authentication. All messages between peers are sent over this protected channel and therefore the peers can verify the data origin of the last hop peer for requests and responses without further cryptography.
In some situations, however, it is desirable to be able to establish the identity of a peer with whom one is not directly connected. The most natural case is when a peer Updates its state. At this point, other peers may need to update their view of the overlay structure, but they need to verify that the Update message came from the actual peer rather than from an attacker. To prevent this, all overlay routing messages are signed by the peer that generated them.
Replay is typically prevented for messages that impact the topology of the overlay by having the information come directly, or be verified by, the nodes that claimed to have generated the update. Data storage replay detection is done by signing time of the node that generated the signature on the store request thus providing a time based replay protection but the time synchronization is only needed between peers that can write to the same location.
The goal here is to stop an attacker from knowing who is signaling what to whom. An attacker is unlikely to be able to observe the activities of a specific individual given the randomization of IDs and routing based on the present peers discussed above. Furthermore, because messages can be routed using only the header information, the actual body of the RELOAD message can be encrypted during transmission.
There are two lines of defense here. The first is the use of TLS or DTLS for each communications link between peers. This provides protection against attackers who are not members of the overlay. The second line of defense is to digitally sign each message. This prevents adversarial peers from modifying messages in flight, even if they are on the routing path.
The routing security mechanisms in RELOAD are designed to contain rather than eliminate attacks on routing. It is still possible for an attacker to mount a variety of attacks. In particular, if an attacker is able to take up a position on the overlay routing between A and B it can make it appear as if B does not exist or is disconnected. It can also advertise false network metrics in an attempt to reroute traffic. However, these are primarily DOS attacks.
The certificate-based security scheme secures the namespace, but if an individual peer is compromised or if an attacker obtains a certificate from the CA, then a number of subversive peers can still appear in the overlay. While these peers cannot falsify responses to resource queries, they can respond with error messages, effecting a DoS attack on the resource registration. They can also subvert routing to other compromised peers. To defend against such attacks, a resource search must still consist of parallel searches for replicated registrations.
This section contains the new code points registered by this document. [NOTE TO IANA/RFC-EDITOR: Please replace RFC-AAAA with the RFC number for this specification in the following list.]
IANA will make the following "Well Known URI" registration as described in [RFC5785]:
[[Note to RFC Editor - this paragraph can be removed before publication. ]] A review request was sent to wellknown-uri-review@ietf.org on October 12, 2010.
| URI suffix: | p2psip-enroll |
| Change controller: | IETF <iesg@ietf.org> |
| Specification document(s): | [RFC-AAAA] |
| Related information: | None |
[[Note to RFC Editor - this paragraph can be removed before publication. ]] IANA has already allocated a TCP port for the main peer to peer protocol. This port has the name p2p-sip and the port number of 6084. IANA needs to update this registration to be defined for UDP as well as TCP.
IANA will make the following port registration:
| Registration Technical Contact | Cullen Jennings <fluffy@cisco.com> |
| Registration Owner | IETF <iesg@ietf.org> |
| Transport Protocol | TCP & UDP |
| Port Number | 6084 |
| Service Name | p2psip-enroll |
| Description | Peer to Peer Infrastructure Enrollment |
| Reference | [RFC-AAAA] |
IANA SHALL create a "RELOAD Overlay Algorithm Type" Registry. Entries in this registry are strings denoting the names of overlay algorithms. The registration policy for this registry is RFC 5226 IETF Review. The initial contents of this registry are:
| Algorithm Name | RFC |
|---|---|
| CHORD-RELOAD | RFC-AAAA |
IANA SHALL create a "RELOAD Access Control Policy" Registry. Entries in this registry are strings denoting access control policies, as described in Section 6.3. New entries in this registry SHALL be registered via RFC 5226 Standards Action. The initial contents of this registry are:
| Access Policy | RFC |
|---|---|
| USER-MATCH | RFC-AAAA |
| NODE-MATCH | RFC-AAAA |
| USER-NODE-MATCH | RFC-AAAA |
| NODE-MULTIPLE | RFC-AAAA |
IANA SHALL create a "RELOAD Application-ID" Registry. Entries in this registry are 16-bit integers denoting application kinds. Code points in the range 0x0001 to 0x7fff SHALL be registered via RFC 5226 Standards Action. Code points in the range 0x8000 to 0xf000 SHALL be registered via RFC 5226 Expert Review. Code points in the range 0xf001 to 0xfffe are reserved for private use. The initial contents of this registry are:
| Application | Application-ID | Specification |
|---|---|---|
| INVALID | 0 | RFC-AAAA |
| SIP | 5060 | Reserved for use by SIP Usage |
| SIP | 5061 | Reserved for use by SIP Usage |
| Reserved | 0xffff | RFC-AAAA |
IANA SHALL create a "RELOAD Data Kind-ID" Registry. Entries in this registry are 32-bit integers denoting data kinds, as described in Section 4.2. Code points in the range 0x00000001 to 0x7fffffff SHALL be registered via RFC 5226 Standards Action. Code points in the range 0x8000000 to 0xf0000000 SHALL be registered via RFC 5226 Expert Review. Code points in the range 0xf0000001 to 0xfffffffe are reserved for private use via the kind description mechanism described in Section 10. The initial contents of this registry are:
| Kind | Kind-ID | RFC |
|---|---|---|
| INVALID | 0 | RFC-AAAA |
| TURN_SERVICE | 2 | RFC-AAAA |
| CERTIFICATE_BY_NODE | 3 | RFC-AAAA |
| CERTIFICATE_BY_USER | 16 | RFC-AAAA |
| Reserved | 0x7fffffff | RFC-AAAA |
| Reserved | 0xfffffffe | RFC-AAAA |
IANA SHALL create a "RELOAD Data Model" Registry. Entries in this registry denoting data models, as described in Section 6.2. Code points in this registry SHALL be registered via RFC 5226 Standards Action. The initial contents of this registry are:
| Data Model | RFC |
|---|---|
| INVALID | RFC-AAAA |
| SINGLE | RFC-AAAA |
| ARRAY | RFC-AAAA |
| DICTIONARY | RFC-AAAA |
| RESERVED | RFC-AAAA |
IANA SHALL create a "RELOAD Message Code" Registry. Entries in this registry are 16-bit integers denoting method codes as described in Section 5.3.3. These codes SHALL be registered via RFC 5226 Standards Action. The initial contents of this registry are:
| Message Code Name | Code Value | RFC |
|---|---|---|
| invalid | 0 | RFC-AAAA |
| probe_req | 1 | RFC-AAAA |
| probe_ans | 2 | RFC-AAAA |
| attach_req | 3 | RFC-AAAA |
| attach_ans | 4 | RFC-AAAA |
| unused | 5 | |
| unused | 6 | |
| store_req | 7 | RFC-AAAA |
| store_ans | 8 | RFC-AAAA |
| fetch_req | 9 | RFC-AAAA |
| fetch_ans | 10 | RFC-AAAA |
| unused (was remove_req) | 11 | RFC-AAAA |
| unused (was remove_ans) | 12 | RFC-AAAA |
| find_req | 13 | RFC-AAAA |
| find_ans | 14 | RFC-AAAA |
| join_req | 15 | RFC-AAAA |
| join_ans | 16 | RFC-AAAA |
| leave_req | 17 | RFC-AAAA |
| leave_ans | 18 | RFC-AAAA |
| update_req | 19 | RFC-AAAA |
| update_ans | 20 | RFC-AAAA |
| route_query_req | 21 | RFC-AAAA |
| route_query_ans | 22 | RFC-AAAA |
| ping_req | 23 | RFC-AAAA |
| ping_ans | 24 | RFC-AAAA |
| stat_req | 25 | RFC-AAAA |
| stat_ans | 26 | RFC-AAAA |
| unused (was attachlite_req) | 27 | RFC-AAAA |
| unused (was attachlite_ans) | 28 | RFC-AAAA |
| app_attach_req | 29 | RFC-AAAA |
| app_attach_ans | 30 | RFC-AAAA |
| unused (was app_attachlite_req) | 31 | RFC-AAAA |
| unused (was app_attachlite_ans) | 32 | RFC-AAAA |
| config_update_req | 33 | RFC-AAAA |
| config_update_ans | 34 | RFC-AAAA |
| reserved | 0x8000..0xfffe | RFC-AAAA |
| error | 0xffff | RFC-AAAA |
IANA SHALL create a "RELOAD Error Code" Registry. Entries in this registry are 16-bit integers denoting error codes. New entries SHALL be defined via RFC 5226 Standards Action. The initial contents of this registry are:
| Error Code Name | Code Value | RFC |
|---|---|---|
| invalid | 0 | RFC-AAAA |
| Unused | 1 | RFC-AAAA |
| Error_Forbidden | 2 | RFC-AAAA |
| Error_Not_Found | 3 | RFC-AAAA |
| Error_Request_Timeout | 4 | RFC-AAAA |
| Error_Generation_Counter_Too_Low | 5 | RFC-AAAA |
| Error_Incompatible_with_Overlay | 6 | RFC-AAAA |
| Error_Unsupported_Forwarding_Option | 7 | RFC-AAAA |
| Error_Data_Too_Large | 8 | RFC-AAAA |
| Error_Data_Too_Old | 9 | RFC-AAAA |
| Error_TTL_Exceeded | 10 | RFC-AAAA |
| Error_Message_Too_Large | 11 | RFC-AAAA |
| Error_Unknown_Kind | 12 | RFC-AAAA |
| Error_Unknown_Extension | 13 | RFC-AAAA |
| Error_Response_Too_Large | 14 | RFC-AAAA |
| Error_Config_Too_Old | 15 | RFC-AAAA |
| Error_Config_Too_New | 16 | RFC-AAAA |
| Error_In_Progress | 17 | RFC-AAAA |
| reserved | 0x8000..0xfffe | RFC-AAAA |
IANA shall create a "RELOAD Overlay Link." New entries SHALL be defined via RFC 5226 Standards Action. This registry SHALL be initially populated with the following values:
| Protocol | Code | Specification |
|---|---|---|
| reserved | 0 | RFC-AAAA |
| DTLS-UDP-SR | 1 | RFC-AAAA |
| DTLS-UDP-SR-NO-ICE | 3 | RFC-AAAA |
| TLS-TCP-FH-NO-ICE | 4 | RFC-AAAA |
| reserved | 255 | RFC-AAAA |
IANA shall create an "Overlay Link Protocol Registry". Entries in this registry SHALL be defined via RFC 5226 Standards Action. This registry SHALL be initially populated with the following value: "TLS".
IANA shall create a "Forwarding Option Registry". Entries in this registry between 1 and 127 SHALL be defined via RFC 5226 Standards Action. Entries in this registry between 128 and 254 SHALL be defined via RFC 5226 Specification Required. This registry SHALL be initially populated with the following values:
| Forwarding Option | Code | Specification |
|---|---|---|
| invalid | 0 | RFC-AAAA |
| reserved | 255 | RFC-AAAA |
IANA shall create a "RELOAD Probe Information Type Registry". Entries in this registry SHALL be defined via RFC 5226 Standards Action. This registry SHALL be initially populated with the following values:
| Probe Option | Code | Specification |
|---|---|---|
| invalid | 0 | RFC-AAAA |
| responsible_set | 1 | RFC-AAAA |
| num_resources | 2 | RFC-AAAA |
| uptime | 3 | RFC-AAAA |
| reserved | 255 | RFC-AAAA |
IANA shall create a "RELOAD Extensions Registry". Entries in this registry SHALL be defined via RFC 5226 Specification Required. This registry SHALL be initially populated with the following values:
| Extensions Name | Code | Specification |
|---|---|---|
| invalid | 0 | RFC-AAAA |
| reserved | 0xFFFF | RFC-AAAA |
This section describes the scheme for a reload URI, which can be used to refer to either:
The reload URI is defined using a subset of the URI schema specified in Appendix A of RFC 3986 [RFC3986] and the associated URI Guidelines [RFC4395] per the following ABNF syntax:
RELOAD-URI = "reload://" destination "@" overlay "/"
[specifier]
destination = 1 * HEXDIG
overlay = reg-name
specifier = 1*HEXDIG
The definitions of these productions are as follows:
If no specifier is present then this URI addresses the peer which can be reached via the indicated destination list at the indicated overlay name. If a specifier is present, then the URI addresses the data value.
[[ Note to RFC Editor - please remove this paragraph before publication. ]] Review request was sent to uri-review@ietf.org on Oct 7, 2010.
The following summarizes the information necessary to register the reload URI.
This specification is a merge of the "REsource LOcation And Discovery (RELOAD)" draft by David A. Bryan, Marcia Zangrilli and Bruce B. Lowekamp, the "Address Settlement by Peer to Peer" draft by Cullen Jennings, Jonathan Rosenberg, and Eric Rescorla, the "Security Extensions for RELOAD" draft by Bruce B. Lowekamp and James Deverick, the "A Chord-based DHT for Resource Lookup in P2PSIP" by Marcia Zangrilli and David A. Bryan, and the Peer-to-Peer Protocol (P2PP) draft by Salman A. Baset, Henning Schulzrinne, and Marcin Matuszewski. Thanks to the authors of RFC 5389 for text included from that. Vidya Narayanan provided many comments and improvements.
The ideas and text for the Chord specific extension data to the Leave mechanisms was provided by J. Maenpaa, G. Camarillo, and J. Hautakorpi.
Thanks to the many people who contributed including Ted Hardie, Michael Chen, Dan York, Das Saumitra, Lyndsay Campbell, Brian Rosen, David Bryan, Dave Craig, and Julian Cain. Extensive working last call comments were provided by: Jouni Maenpaa, Roni Even, Ari Keranen, John Buford, Michaelx Chen, Frederic-Philippe Met, and David Bryan. Special thanks to Marc Petit-Huguenin who provied an amazing amount to detailed review.
Significant discussion has been focused on the selection of a routing algorithm for P2PSIP. This section discusses the motivations for selecting symmetric recursive routing for RELOAD and describes the extensions that would be required to support additional routing algorithms.
Iterative routing has a number of advantages. It is easier to debug, consumes fewer resources on intermediate peers, and allows the querying peer to identify and route around misbehaving peers [non-transitive-dhts-worlds05]. However, in the presence of NATs, iterative routing is intolerably expensive because a new connection must be established for each hop (using ICE) [bryan-design-hotp2p08].
Iterative routing is supported through the RouteQuery mechanism and is primarily intended for debugging. It also allows the querying peer to evaluate the routing decisions made by the peers at each hop, consider alternatives, and perhaps detect at what point the forwarding path fails.
An alternative to the symmetric recursive routing method used by RELOAD is Forward-Only routing, where the response is routed to the requester as if it were a new message initiated by the responder (in the previous example, Z sends the response to A as if it were sending a request). Forward-only routing requires no state in either the message or intermediate peers.
The drawback of forward-only routing is that it does not work when the overlay is unstable. For example, if A is in the process of joining the overlay and is sending a Join request to Z, it is not yet reachable via forward routing. Even if it is established in the overlay, if network failures produce temporary instability, A may not be reachable (and may be trying to stabilize its network connectivity via Attach messages).
Furthermore, forward-only responses are less likely to reach the querying peer than symmetric recursive ones are, because the forward path is more likely to have a failed peer than is the request path (which was just tested to route the request) [non-transitive-dhts-worlds05].
An extension to RELOAD that supports forward-only routing but relies on symmetric responses as a fallback would be possible, but due to the complexities of determining when to use forward-only and when to fallback to symmetric, we have chosen not to include it as an option at this point.
Another routing option is Direct Response routing, in which the response is returned directly to the querying node. In the previous example, if A encodes its IP address in the request, then Z can simply deliver the response directly to A. In the absence of NATs or other connectivity issues, this is the optimal routing technique.
The challenge of implementing direct response is the presence of NATs. There are a number of complexities that must be addressed. In this discussion, we will continue our assumption that A issued the request and Z is generating the response.
An extension to RELOAD that supports direct response routing but relies on symmetric responses as a fallback would be possible, but due to the complexities of determining when to use direct response and when to fallback to symmetric, and the reduced performance for responses to peers behind restrictive NATs, we have chosen not to include it as an option at this point.
[I-D.jiang-p2psip-relay] has proposed implementing a form of direct response by having A identify a peer, Q, that will be directly reachable by any other peer. A uses Attach to establish a connection with Q and advertises Q's IP address in the request sent to Z. Z sends the response to Q, which relays it to A. This then reduces the latency to two hops, plus Z negotiating a secure connection to Q.
This technique relies on the relative population of nodes such as A that require relay peers and peers such as Q that are capable of serving as a relay peer. It also requires nodes to be able to identify which category they are in. This identification problem has turned out to be hard to solve and is still an open area of exploration.
An extension to RELOAD that supports relay peers is possible, but due to the complexities of implementing such an alternative, we have not added such a feature to RELOAD at this point.
A concept similar to relay peers, essentially choosing a relay peer at random, has previously been suggested to solve problems of pairwise non-transitivity [non-transitive-dhts-worlds05], but deterministic filtering provided by NATs makes random relay peers no more likely to work than the responding peer.
A common concern about symmetric recursive routing has been that one or more peers along the request path may fail before the response is received. The significance of this problem essentially depends on the response latency of the overlay. An overlay that produces slow responses will be vulnerable to churn, whereas responses that are delivered very quickly are vulnerable only to failures that occur over that small interval.
The other aspect of this issue is whether the request itself can be successfully delivered. Assuming typical connection maintenance intervals, the time period between the last maintenance and the request being sent will be orders of magnitude greater than the delay between the request being forwarded and the response being received. Therefore, if the path was stable enough to be available to route the request, it is almost certainly going to remain available to route the response.
An overlay that is unstable enough to suffer this type of failure frequently is unlikely to be able to support reliable functionality regardless of the routing mechanism. However, regardless of the stability of the return path, studies show that in the event of high churn, iterative routing is a better solution to ensure request completion [lookups-churn-p2p06] [non-transitive-dhts-worlds05]
Finally, because RELOAD retries the end-to-end request, that retry will address the issues of churn that remain.
There are a wide variety of reasons a node may act as a client rather than as a peer [I-D.pascual-p2psip-clients]. This section outlines some of those scenarios and how the client's behavior changes based on its capabilities.
For a number of reasons, a particular node may be forced to act as a client even though it is willing to act as a peer. These include:
The ultimate criteria for a node to become a peer are determined by the overlay algorithm and specific deployment. A node acting as a client that has a full implementation of RELOAD and the appropriate overlay algorithm is capable of locating its responsible peer in the overlay and using Attach to establish a direct connection to that peer. In that way, it may elect to be reachable under either of the routing approaches listed above. Particularly for overlay algorithms that elect nodes to serve as peers based on trustworthiness or population, the overlay algorithm may require such a client to locate itself at a particular place in the overlay.
SIP defines an extensive protocol for registration and security between a client and its registrar/proxy server(s). Any SIP device can act as a client of a RELOAD-based P2PSIP overlay if it contacts a peer that implements the server-side functionality required by the SIP protocol. In this case, the peer would be acting as if it were the user's peer, and would need the appropriate credentials for that user.
Application-level support for clients is defined by a usage. A usage offering support for application-level clients should specify how the security of the system is maintained when the data is moved between the application and RELOAD layers.