| TCPM Working Group | O. Bonaventure, Ed. |
| Internet-Draft | Tessares |
| Intended status: Experimental | M. Boucadair, Ed. |
| Expires: September 6, 2018 | Orange |
| B. Peirens | |
| Proximus | |
| S. Seo | |
| Korea Telecom | |
| A. Nandugudi | |
| Memphis University | |
| March 05, 2018 |
0-RTT TCP Convert Protocol
draft-ietf-tcpm-converters-01
This document specifies an application proxy, called Transport Converter, to assist the deployment of TCP extensions such as Multipath TCP. This proxy is designed to avoid inducing extra delay when involved in a network-assisted connection (that is, 0-RTT). This specification assumes an explicit model, where the proxy is explicitly configured on hosts.
— Editorial Note (To be removed by RFC Editor)
Please update these statements with the RFC number to be assigned to this document:
[This-RFC]
Please update TBA statements with the port number to be assigned to the Converter Protocol.
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 https://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 6, 2018.
Copyright (c) 2018 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 (https://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.
Transport protocols like TCP evolve regularly [RFC7414]. TCP has been improved in different ways. Some improvements such as changing the initial window size [RFC6928] or modifying the congestion control scheme can be applied independently on clients and servers. Other improvements such as Selective Acknowledgements [RFC2018] or large windows [RFC7323] require a new TCP option or to change the semantics of some fields in the TCP header. These modifications must be deployed on both clients and servers to be actually used on the Internet. Experience with the latter TCP extensions reveals that their deployment can require many years. [Fukuda2011] reports results of a decade of measurements showing the deployment of Selective Acknowledgements, Window Scale and TCP Timestamps. [ANRW17] describes measurements showing that TCP Fast Open [RFC7413] (TFO) is still not widely deployed.
There are some situations where the transport stack used on clients (resp. servers) can be upgraded at a faster pace than the transport stack running on servers (resp. clients). In those situations, clients would typically want to benefit from the features of an improved transport protocol even if the servers have not yet been upgraded and conversely. In the past, Performance Enhancing Proxies have been proposed and deployed [RFC3135] as solutions to improve TCP performance over links with specific characteristics.
Recent examples of TCP extensions include Multipath TCP [RFC6824][I-D.ietf-mptcp-rfc6824bis] or TCPINC [I-D.ietf-tcpinc-tcpcrypt]. Those extensions provide features that are interesting for clients such as wireless devices. With Multipath TCP, those devices could seamlessly use WLAN and cellular networks, for bonding purposes, faster handovers, or better resiliency. Unfortunately, deploying those extensions on both a wide range of clients and servers remains difficult.
More recently, experimentation of 5G bonding, which has very scarce coverage, has been conducted into global range of the incumbent 4G (LTE) connectivity in newly devised clients using Multipath TCP proxy. Even if the 5G and the 4G bonding by using Multipath TCP increases the bandwidth to data transfer, it is as well crucial to minimize latency for all the way between endhosts regardless of whether intermediate nodes are inside or ouside of the mobile core. In order to handle uRLLC (Ultra-Reliable Low-Latency Communication) for the next generation mobile network, Multipath TCP and its proxy mechanism must be optimized to reduce latency.
This document specifies an application proxy, called Transport Converter. A Transport Converter is a function that is installed by a network operator to aid the deployment of TCP extensions and to provide the benefits of such extensions to clients. A Transport Converter may support conversion service for one or more TCP extensions. This service is provided by means of the Converter Protocol (Convert), that is an application layer protocol which uses TBA TCP port number (Section 8).
The Transport Converter adheres to the main principles as drawn in [RFC1919]. In particular, the Converter achieves the following:
The main advantage of network-assisted Converters is that they enable new TCP extensions to be used on a subset of the end-to-end path, which encourages the deployment of these extensions. The Transport Converter allows the client and the server to directly negotiate TCP options.
The Convert Protocol is a generic mechanism to provide 0-RTT conversion service. As a sample applicability use case, this document specifies how the Convert Protocol applies for Multiptah TCP. It is out of scope of this document to provide a comprehensive list of potential all conversion services; separate documents may be edited in the future for other conversion services upon need.
This document does not assume that all the traffic is eligible to the network-assisted conversion service. Only a subset of the traffic will be forwarded to a Converter according to a set of policies. Furthermore, it is possible to bypass the Converter to connect to the servers that already support the required TCP extension.
This document assumes that a client is configured with one or a list of Converters. Configuration means are outside the scope of this document.
This document is organized as follows. We first provide a brief explanation of the operation of Transport Converters in Section 3. We describe the Converter Protocol in Section 4. We discuss in Section 5 how Transport Converters can be used to support different TCP options. We then discuss the interactions with middleboxes (Section 6) and the security considerations (Section 7).
Appendix A provides a comparison with SOCKS proxies that are already used to deploy Multipath TCP in some cellular networks.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in [RFC2119] [RFC8174] when, and only when, they appear in all capitals,
as shown here.
The architecture considers three types of endhosts:
A Transport Converter is a network function that relays all data exchanged over one upstream connection to one downstream connection and vice versa (Figure 1). The Converter, thus, maintains state that associates one upstream connection to a corresponding downstream connection.
A connection can be initiated from both sides of the Transport Converter (Internet-facing interface, client-facing interface).
+------------+
<--- upstream --->| Transport |<--- downstream --->
| Converter |
+------------+
Figure 1: A Transport Converter relays data between pairs of TCP connections
Transport Converters can be operated by network operators or third parties. Nevertheless, this document focuses on the single administrative deployment case where the entity offering the connectivity service to a client is also the entity which owns and operates the Transport Converter.
A Transport Converter can be embedded in a standalone device or be activated as a service on a router. How such function is enabled is deployment-specific (Figure 2).
+-+ +-+ +-+
Client - |R| -- |R| -- |R| - - - Server
+-+ +-+ +-+
|
Transport
Converter
Figure 2: A Transport Converter can be installed anywhere in the network
The architecture assumes that new software will be installed on the Client hosts and on Transport Converters. Further, the architecture allows for making use of TCP new extensions if those are supported by a given server.
The Client is configured, through means that are outside the scope of this document, with the names and/or the addresses of one or more Transport Converters.
The architecture does not mandate anything on the server side.
One of the benefits of this design is that different transport protocol extensions can be used on the upstream and the downstream connections. This encourages the deployment of new TCP extensions until they are widely supported by servers.
At a high level, the objective of the Transport Converter is to allow the Client to use a specific extension, e.g. Multipath TCP, on a subset of the end-to-end path even if the Server does not support this extension. This is illustrated in Figure 3 where the Client initiates a Multipath TCP connection with the Converter (Multipath packets are shown with "===") while the Converter uses a regular TCP connection with the Server.
The packets belonging to the pair of connections between the Client and Server passing through a Transport Converter may follow a different path than the packets directly exchanged between the Client and the Server. Deployments should minimize the possible additional delay by carefully selecting the location of the Transport Converter used to reach a given destination.
Transport
Client Converter Server
======================>
-------------------->
<--------------------
<======================
Multipath TCP packets Regular TCP packets
Figure 3: Different TCP variants can be used on the Client-Converter path and on the Converter-Server path
When establishing a connection, the Client can, depending on local policies, either contact the Server directly (e.g., by sending a TCP SYN towards the Server) or create the connection via a Transport Converter. In the latter case, which is the case we consider in this document, the Client initiates a connection towards the Transport Converter and indicates the IP address and port number of the ultimate Server inside the connection establishment packet. Doing so enables the Transport Converter to immediately initiate a connection towards that Server, without experiencing an extra delay. The Transport Converter waits until the confirmation that the Server agrees to establish the connection before confirming it to the Client.
The client places the destination address and port number of the target Server in the payload of the SYN sent to the Converter by leveraging TCP Fast Open [RFC7413]. In accordance with [RFC1919], the Transport Converter maintains two connections that are combined together:
Any user data received by the Transport Converter over the upstream (resp., downstream) connection is relayed over the downstream (resp., upstream) connection.
Figure 4 illustrates the establishment of a TCP connection by the Client through a Transport Converter. The information shown between brackets is part of the Converter Protocol described later in this document.
Transport
Client Converter Server
-------------------->
SYN TFO [->Server:port]
-------------------->
SYN
<--------------------
SYN+ACK
<--------------------
SYN+ACK [ ]
Figure 4: Establishment of a TCP connection through a Converter
The Client sends a SYN destined to the Transport Converter. This SYN contains a TFO cookie and inside its payload the addresses and ports of the destination Server. The Transport Converter does not reply immediately to this SYN. It first tries to create a TCP connection towards the destination Server. If this second connection succeeds, the Transport Converter confirms the establishment of
the connection to the Client by returning a SYN+ACK and the first bytes of the bytestream contain information about the TCP options that were negotiated with the final Server. This information is sent at the beginning of the bytestream, either directly in the SYN+ACK or in a subsequent packet. For graphical reasons, the figures in this section show that the Converter returns this information in the SYN+ACK packet. An implementation could also place this information in a packet that it sent shortly after the SYN+ACK.
The connection can also be established from the Internet towards a Client via a Transport Converter. This is typically the case when the Client embeds a server (video server, for example).
The procedure described in Figure 4 assumes that the Client has obtained a TFO cookie from the Transport Converter. This is part of the Bootstrap procedure which is illustrated in Figure 5. The Client sends a SYN with a TFO request option to obtain a valid cookie from the Converter. The Converter replies with a TFO cookie in the SYN+ACK. Once this connection has been established, the Client sends a Bootstrap message to request the list of TCP options for which the Transport Converter provides a conversion service.
Transport
Client Converter Server
-------------------->
SYN TFO(empty)
<--------------------
SYN+ACK TFO(cookie)
-------------------->
[Bootstrap]
<--------------------
[Supported TCP Extension Services]
Figure 5: Bootstrapping a Client connection to a Transport Converter
Note that the Converter may rely on local policies to decide whether it can service a given requesting Client. That is, the Converter will not return a cookie for that Client. How such policies are supplied to the Converter are out of scope.
Also, the Converter may behave in a cookie-less mode when appropriate means are enforced at the Converter and the network in-between to protect against attacks such as spoofing and SYN flood. Under such deployments, the use of TFO is not required.
As an example (Figure 6), let us consider how the Convert protocol can help the deployment of Multipath TCP [RFC6824]. We assume that both the Client and the Transport Converter support Multipath TCP, but consider two different cases depending whether the Server supports Multipath TCP or not. A Multipath TCP connection is created by placing the MP_CAPABLE (MPC) option in the SYN sent by the Client.
Figure 6 describes the operation of the Transport Converter if the Server does not support Multipath TCP.
Transport
Client Converter Server
-------------------->
SYN, MPC [->Server:port]
-------------------->
SYN, MPC
<--------------------
SYN+ACK
<--------------------
SYN+ACK,MPC [ ]
-------------------->
ACK,MPC
-------------------->
ACK
Figure 6: Establishment of a Multipath TCP connection through a Converter
The Client tries to initiate a Multipath TCP connection by sending a SYN with the MP_CAPABLE option (MPC in Figure 6). The SYN includes the address and port number of the final Server and the Transport Converter attempts to initiate a Multipath TCP connection towards this Server. Since the Server does not support Multipath TCP, it replies with a SYN+ACK that does not contain the MP_CAPABLE option. The Transport Converter notes that the connection with the Server does not support Multipath TCP and returns the TCP options received from the Server to the Client.
Figure 7 considers a Server that supports Multipath TCP. In this case, it replies to the SYN sent by the Transport Converter with the MP_CAPABLE option. Upon reception of this SYN+ACK, the Transport Converter confirms the establishment of the connection to the Client and indicates to the Client that the Server supports Multipath TCP. With this information, the Client has discovered that the Server supports Multipath TCP natively. This will enable it to bypass the Transport Converter for the next Multipath TCP connection that it will initiate towards this Server.
Transport
Client Converter Server
-------------------->
SYN, MPC [->Server:port]
-------------------->
SYN, MPC
<--------------------
SYN+ACK, MPC
<--------------------
SYN+ACK, MPC [ MPC supported ]
-------------------->
ACK, MPC
-------------------->
ACK, MPC
Figure 7: Establishment of a Multipath TCP connection through a converter
An example of an incoming Converter-assisted Multipath TCP connection is depicted in Figure 8. In order to support incoming connections from remote hosts, the Client may use PCP [RFC6887] to instruct the Converter to create dynamic mappings. Those mappings will be used by the Converter to intercept an incoming TCP connection destined to the Client and convert it into a Multipath TCP connection.
Transport
Client Converter Remote Host
<-------------------
SYN
<-------------------
SYN, MPC[Remote Host:port]
--------------------->
SYN+ACK, MPC
--------------------->
SYN+ACK
<---------------------
ACK
<-------------------
ACK, MPC
Figure 8: Establishment of an Incoming TCP Connection through a Converter
This section describes in details the messages that are exchanged between a Client and a Transport Converter. The Converter Protocol (Convert, for short) leverages the TCP Fast Open extension [RFC7413].
The Converter Protocol uses a 32 bits long fixed header that is sent by both the Client and the Transport Converter. This header indicates both the version of the protocol used and the length of the Convert message.
The Fixed Header is used to exchange information about the version and length of the messages between the Client and the Transport Converter.
The Client and the Transport Converter MUST send the fixed-sized header shown in Figure 9 as the first four bytes of the bytestream.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+-------------------------------+
| Version | Total Length | Unassigned |
+---------------+---------------+-------------------------------+
Figure 9: The fixed-sized header of the Converter protocol
The Version is encoded as an 8 bits unsigned integer value. This document specifies version 1. Version 0 is reserved by this document and MUST NOT be used.
The Total Length is the number of 32 bits word, including the header, of the bytestream that are consumed by the Converter protocol messages. Since Total Length is also an 8 bits unsigned integer, those messages cannot consume more than 1020 bytes of data. This limits the number of bytes that a Transport Converter needs to process. A Total Length of zero is invalid and the connection MUST be reset upon reception of such a header.
The Unassigned field MUST be set to zero in this version of the protocol. These bits are available for future use [RFC8126].
The Convert protocol uses variable length messages that are encoded using the generic TLV format depicted in Figure 10. All TLV fields are encoded using the network byte order.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+-------------------------------+
| Type | Length | (optional) Value ... |
+---------------+---------------+-------------------------------+
| ... (optional) Value |
+---------------------------------------------------------------+
Figure 10: Converter Generic TLV Format
A given TLV MUST only appear once on a connection. If two or more instances of the same TLV are exchanged over a Converter connection, the associated TCP connections MUST be closed.
This document specifies the following Convert TLVs:
+------+-----+----------+------------------------------------------+ | Type | Hex | Length | Description | +------+-----+----------+------------------------------------------+ | 1 | 0x1 | 1 | Bootstrap TLV | | 10 | 0xA | Variable| Connect TLV | | 20 | 0x14| Variable| Extended TCP Header TLV | | 21 | 0x15| Variable| Supported TCP Extension Services TLV | | 30 | 0x1E| Variable| Error TLV | +------+-----+----------+------------------------------------------+
Figure 11: The TLVs used by the Converter protocol
To establish a connection via a Transport Converter, a Client MUST first obtain a valid TFO cookie from that Converter. This is the bootstrap procedure during which the Client opens a connection to the Transport Converter with an empty TFO option. According to [RFC7413], the Transport Converter returns its cookie in the SYN+ACK. Then the Client sends a Bootstrap TLV (Section 4.2.3) to which the Transport Converter replies with the Supported TCP Extension Services TLV described in Section 4.2.4.
With the TFO cookie of the Transport Converter, the Client can request the establishment of connections to remote servers with the Connect TLV (see Section 4.2.5). If the connection can be established with the final server, the Transport Converter replies with the Extended TCP Header TLV and returns an Error TLV inside a RST packet (see Section 4.2.7).
When the Transport Converter receives an incoming connection
establishment from a Client, it MUST process the TCP options found in the SYN and the Connect TLV. In general, the Transport Converter MUST add to the proxied SYN the TCP options that were included in the Connect TLV. It SHOULD add to the proxied SYN the TCP options that
were included in the incoming SYN provided that it supports the corresponding TCP extension.
There are some exceptions to these rules given the semantics of some
TCP options. First, TCP options with Kinds 0 (EOL), 1 (NOP), 2
(MSS), and 3 (WS) MUST be used according to the configuration of the
TCP stack of the Transport Converter. The Timestamps option
(Kind=10) SHOULD be used in the proxied SYN if it was present in the
incoming SYN, but the contents of the option in the proxied SYN SHOULD be set by the Converter's stack. The MP_CAPABLE option SHOULD
be added to the proxied SYN if it was present in the incoming SYN,
but the content of the option in the proxied SYN SHOULD be set by the
Converter's stack. The TCP Fast Open cookie option SHOULD be handled
as described in Section 6.
As a general rule, when an error is encountered an Error TLV with the appropriate error code MUST be returned.
The Bootstrap TLV (Figure 12 is sent by a Client to request the TCP extensions that are supported by a Transport Converter and for which it provides a conversion service. It is typically sent on the first connection that a Client establishes with a Transport Converter to learn its capabilities. Assuming a Client is entitled to invoke the Converter, this latter replies with the Supported TCP Extensions Services TLV described in Section 4.2.4.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+-------------------------------+
| Type | Length | Zero |
+---------------+---------------+-------------------------------+
Figure 12: The Bootstrap TLV
The Supported TCP Extension Services TLV (Figure 13) is used by a Converter to announce the TCP options for which it provides a conversion service. Each supported TCP option is encoded with its TCP option Kind listed in the "TCP Parameters" registry maintained by IANA.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+-------------------------------+
| Type | Length | Unassigned |
+---------------+---------------+-------------------------------+
| Kind #1 | Kind #2 | ... |
+---------------+---------------+-------------------------------+
/ ... /
/ /
+---------------------------------------------------------------+
Figure 13: The Supported TCP Extension Services TLV
TCP option Kinds 0, 1, and 2 defined in [RFC0793] are supported by all TCP implementations and thus MUST NOT appear in this list.
The list of Supported TCP Extension Services is padded with 0 to end on a 32 bits boundary.
Typically, if the Converter only supports Multipath TCP conversion service, solely Kind=30 will be present in the Supported TCP Extension Services TLV returned by the Converter to a requesting Client.
The Connect TLV (Figure 14) is used to request the establishment of a connection via a Transport Converter.
The 'Remote Peer Port' and 'Remote Peer IP Address' fields contain the destination port and IP address of the target server for an outgoing connection towards a server located on the Internet. For incoming connections destined to a client serviced via a Converter, these fields convey the source port and IP address.
The Remote Peer IP Address MUST be encoded as an IPv6 address. IPv4 addresses MUST be encoded using the IPv4-Mapped IPv6 Address format defined in [RFC4291].
The optional 'TCP Options' field is used to specify how specific TCP Options should be advertised by the Transport Converter to the final destination of a connection. If this field is not supplied, the Transport Converter MUST use the default TCP options that correspond to its local policy.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+-------------------------------+
| Type | Length | Remote Peer Port |
+---------------+---------------+-------------------------------+
| |
| Remote Peer IP Address (128 bits) |
| |
| |
+---------------------------------------------------------------+
| TCP Options (Variable) |
| ... |
+---------------------------------------------------------------+
Figure 14: The Connect TLV
The 'TCP Options' field is a variable length field that carries a list of TCP option fields (Figure 15). Each TCP option field is encoded as a block of 2+n bytes where the first byte is the TCP option Type and the second byte is the length of the TCP option as specified in [RFC0793]. The minimum value for the TCP option Length is 2. The TCP options that do not include a length subfield, i.e., option types 0 (EOL) and 1 (NOP) defined in [RFC0793] cannot be placed inside the TCP options field of the Connect TLV. The optional Value field contains the variable-length part of the TCP option. A length of two indicates the absence of the Value field. The TCP options field always ends on a 32 bits boundary after being padded with zeros.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+---------------+---------------+
| TCPOpt type | TCPOpt Length | Value (opt) | .... |
+---------------+---------------+---------------+---------------+
| .... |
+---------------------------------------------------------------+
| ... |
+---------------------------------------------------------------+
Figure 15: The TCP Options field
If a Transport Converter receives a Connect TLV with a non-empty TCP options field, and the Converter accets to process the request, it SHALL present those options to the destination peer in addition to the TCP options that it would have used according to its local policies. For the TCP options that are listed without an optional value, the Converter MUST generate its own value. For the TCP options that are included in the 'TCP Options' field with an optional value, it SHALL copy the entire option for use in the connection with the destination peer. This feature is required to support TCP Fast Open.
The Converter may discard a Connect TLV request for many reasons (e.g., bad TFO cookie, authorization failed, out of resources). An error message indicating the encountered error is returned to the requesting Client Section 4.2.7. In order to prevent denial-of-service attacks, error messages sent to a Client SHOULD be rate-limited.
The Extended TCP Header TLV (Figure 16) is used by the Transport Converter to send to the Client the extended TCP header that was returned by the Server in the SYN+ACK packet. This TLV is only sent if the Client sent a Connect TLV to request the establishment of a connection.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+-------------------------------+
| Type | Length | Unassigned |
+---------------+---------------+-------------------------------+
| Returned Extended TCP header |
| ... |
+---------------------------------------------------------------+
Figure 16: The Extended TCP Header TLV
The Returned Extended TCP header field is a copy of the extended header that was received in the SYN+ACK by the Transport Converter.
The Unassigned field MUST be set to zero by the transmitter and ignored by the receiver. These bits are available for future use [RFC8126].
The optional Error TLV (Figure 17) can be used by the Transport Converter to provide information about some errors that occurred during the processing of a request to convert a connection. This TLV appears after the Convert header in a RST segment returned by the Transport Converter if the error is fatal and prevented the establishment of the connection. If the error is not fatal and the connection could be established with the final destination, then the error TLV will be carried in the payload.
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+---------------+---------------+----------------+--------------+
| Type | Length | Error | Value |
+---------------+---------------+----------------+--------------+
Figure 17: The Error TLV
Different types of errors can occur while processing Convert messages. Each error is identified by a code represented as an unsigned integer. Four classes of errors are defined:
The following error codes are defined in this document:
Figure 18 summarizes the different error codes.
+-------+------+-----------------------------------------------+ | Error | Hex | Description | +-------+------+-----------------------------------------------+ | 0 | 0x00 | Unsupported Version | | 1 | 0x01 | Malformed Message | | 2 | 0x02 | Unsupported Message | | 32 | 0x20 | Not Authorized | | 33 | 0x21 | Unsupported TCP Option | | 64 | 0x40 | Resource Exceeded | | 65 | 0x41 | Network Failure | | 96 | 0x60 | Connection Reset | | 97 | 0x61 | Destination Unreachable | +-------+------+-----------------------------------------------+
Figure 18: Convert Error Values
In this section, we discuss how several standard track TCP options can be supported through the Converter. The non-standard track options and the experimental options will be discussed in other documents.
Three TCP options were initially defined in [RFC0793] : End-of-Option List (Kind=0), No-Operation (Kind=1) and Maximum Segment Size (Kind=2). The first two options are mainly used to pad the TCP extended header. There is no reason for a client to request a Converter to specifically send these options towards the final destination.
The Maximum Segment Size option (Kind=2) is used by a host to indicate the largest segment that it can receive over each connection. This value is function of the stack that terminates the TCP connection. There is no reason for a Client to request a Converter to advertise a specific MSS value to a remote server.
A Converter MUST ignore options with Kind=0, 1 or 2 if they appear in a Connect TLV. It MUST NOT announce them in a Bootstrap TLV.
The Window Scale option (Kind=3) is defined in [RFC7323]. As for the MSS option, the window scale factor that is used for a connection strongly depends on the TCP stack that handles the connection. When a Converter opens a TCP connection towards a remote server on behalf of a Client, it SHOULD use a WS option with a scaling factor that corresponds to the configuration of its stack. A local configuration MAY allow for WS option in the proxied message to be function of the scaling factor of the incoming connection.
There is no benefit from a deployment viewpoint in enabling a Client of a Converter to specifically request the utilisation of the WS option (Kind=3) with a specific scaling factor towards a remote Server. For this reason, a Converter MUST ignore option Kind=3 if it appears in a Connect TLV. It MUST NOT announce it in a Bootstrap TLV.
Two distinct TCP options were defined to support selective acknowledgements in [RFC2018]. This first one, SACK Permitted (Kind=4), is used to negotiate the utilisation of selective acknowledgements during the three-way handshake. The second one, SACK (Kind=5), carries the selective acknowledgements inside regular segments.
The SACK Permitted option (Kind=4) MAY be advertised by a Transport Converter in the Bootstrap TLV. In this case, Clients connected to this Transport Converter MAY include the SACK Permitted option in the Connect TLV.
The SACK option (Kind=5) cannot be used during the three-way handshake. For this reason, a Transport Converter MUST ignore option Kind=5 with if it appears in a Connect TLV. It MUST NOT announce it in a Bootstrap TLV.
The Timestamp option was initially defined in [RFC1323] which has been replaced by [RFC7323]. It can be used during the three-way handshake to negotiate the utilisation of the timestamps during the TCP connection. It is notably used to improve round-trip-time estimations and to provide protection against wrapped sequence numbers (PAWS). As for the WS option, the timestamps are a property of a connection and there is limited benefit in enabling a client to request a Converter to use the timestamp option when establishing a connection to a remote server. Furthermore, the timestamps that are used by TCP stacks are specific to each stack and there is no benefit in enabling a client to specify the timestamp value that a Converter could use to establish a connection to a remote server.
A Transport Converter MAY advertise the Timestamp option (Kind=8) in the Bootstrap TLV. The clients connected to this Converter MAY include the Timestamp option in the Connect TLV but without any timestamp.
The Multipath TCP options are defined in [RFC6824]. [RFC6824] defines one variable length TCP option (Kind=30) that includes a subtype field to support several Multipath TCP options. There are several operational use cases where clients would like to use Multipath TCP through a Converter [IETFJ16]. However, none of these use cases require the Client to specify the content of the Multipath TCP option that the Converter should send to a remote server.
A Transport Converter which supports Multipath TCP conversion service MUST advertise the Multipath TCP option (Kind=30) in the Bootstrap TLV. Clients serviced by this Converter may include the Multipath TCP option in the Connect TLV but without any content.
The TCP Fast Open cookie option (Kind=34) is defined in [RFC7413]. There are two different usages of this option that need to be supported by Transport Converters. The first utilisation of the Fast Open cookie is to request a cookie from the server. In this case, the option is sent with an empty cookie by the client and the server returns the cookie. The second utilisation of the Fast Open cookie is to send a cookie to the server. In this case, the option contains a cookie.
A Transport Converter MAY advertise the TCP Fast Open cookie option (Kind=34) in the Bootstrap TLV. If a Transport Converter has advertised the support for TCP Fast Open in its Bootstrap TLV, it needs to be able to process two types of Connect TLV. If such a Transport Converter receives a Connect TLV with the TCP Fast Open cookie option that does not contain a cookie, it MUST add an empty TCP Fast Open cookie option in the SYN sent to the remote server. If such a Transport Converter receives a Connect TLV with the TCP Fast Open cookie option that contains a cookie, it MUST copy the TCP Fast Open cookie option in the SYN sent to the remote server.
The TCP User Timeout option is defined in [RFC5482]. The associated TCP option (Kind=28) does not appear to be widely deployed.
Editor's Note: Feedback requested for the utilisation of this option by deployed TCP stacks.
TCP-AO [RFC5925] provides a technique to authenticate all the packets exchanged over a TCP connection. Given the nature of this extension, it is unlikely that the applications that require their packets to be authenticated end-to-end would want their connections to pass through a converter. For this reason, we do not recommend the support of the TCP-AO option by Transport Converters. The only use cases where is makes sense to combine TCP-AO and the solution in this document are those where the TCP-AO-NAT extension [RFC6978] is in use.
A Converter MUST NOT advertise the TCP-AO option (Kind=29) in the Bootstrap TLV. If a Converter receives a Connect TLV that contains the TCP-AO option, it MUST reject the establishment of the connection with error code set to "Unsupported TCP Option", except if the TCP-AO-NAT option is used.
The TCP Experimental options are defined in [RFC4727]. Given the variety of semantics for these options and their experimental nature, it is impossible to discuss them in details in this document.
The Converter Protocol was designed to be used in networks that do not contain middleboxes that interfere with TCP. We describe in this section how a Client can detect middlebox interference and stop using the Transport Converter affected by this interference.
Internet measurements [IMC11] have shown that middleboxes can affect the deployment of TCP extensions. In this section, we only discuss the middleboxes that modify SYN and SYN+ACK packets since the Converter Protocol places its messages in such packets.
Let us first consider a middlebox that removes the TFO Option from the SYN packet. This interference will be detected by the Client during the bootstrap procedure discussed in Section 4.2.3. A Client should not use a Transport Converter that does not reply with the TFO option during the Bootstrap.
Consider a middlebox that removes the SYN payload after the bootstrap procedure. The Client can detect this problem by looking at the acknowledgement number field of the SYN+ACK returned by the Transport Converter. The Client should stop to use this Transport Converter given the middlebox interference.
As explained in [RFC7413], some carrier-grade NATs can affect the operation of TFO if they assign different IP addresses to the same end host. Such carrier-grade NATs could affect the operation of the TFO Option used by the Converter Protocol. See also the discussion in Section 7.1 of [RFC7413].
The Converter may have access to privacy-related information (e.g., subscriber credentials). The Converter MUST NOT leak such sensitive information outside a local domain.
Given its function and its location in the network, a Transport Converter has access to the payload of all the packets that it processes. As such, it MUST be protected as a core IP router (e.g., [RFC1812]).
Furthermore, ingress filtering policies MUST be enforced at the network boundaries [RFC2827].
This document assumes that all network attachments are managed by the same administrative entity. Therefore, enforcing anti-spoofing filters at these network ensures that hosts are not sending traffic with spoofed source IP addresses.
The Converter Protocol is intended to be used in managed networks where end hosts can be identified by their IP address. Thanks to the Bootstrap procedure, the Transport Converter can verify that the Client correctly receives packets sent by the Converter. Stronger authentication schemes MUST be defined to use the Converter Protocol in more open network environments; such schemes are out of scope of this document.
See below for authorization considerations that are specific for Multipath TCP.
Another possible risk is the amplification attacks since a Transport Converter sends a SYN towards a remote Server upon reception of a SYN from a Client. This could lead to amplification attacks if the SYN sent by the Transport Converter were larger than the SYN received from the Client or if the Transport Converter retransmits the SYN. To mitigate such attacks, the Transport Converter SHOULD rate limit the number of pending requests for a given Client. It SHOULD also avoid sending to remote Servers SYNs that are significantly longer than the SYN received from the Client. In practice, Transport Converters SHOULD NOT advertise to a Server TCP options that were not specified by the Client in the received SYN. Finally, the Transport Converter SHOULD only retransmit a SYN to a Server after having received a retransmitted SYN from the corresponding Client.
Upon reception of a SYN that contains a valid TFO cookie and a Connect TLV, the Transport Converter attempts to establish a TCP connection to a remote Server. There is a risk of denial of service attack if a Client requests too many connections in a short period of time. Implementations SHOULD limit the number of pending connections from a given Client. Means to protect against SYN flooding attacks MUST also be enabled [RFC4987].
Traffic theft is a risk if an illegitimate Converter is inserted in the path. Indeed, inserting an illegitimate Converter in the forwarding path allows traffic interception and can therefore provide access to sensitive data issued by or destined to a host. Converter discovery and configuration are out of scope of this document.
Multipath TCP-related security threats are discussed in [RFC6181] and [RFC6824].
The operator that manages the various network attachments (including the Converters) can enforce authentication and authorization policies using appropriate mechanisms. For example, a non-exhaustive list of methods to achieve authorization is provided hereafter:
A first safeguard against the misuse of Converter resources by illegitimate users (e.g., users with access networks that are not managed by the same provider that operates the Converter) is the Converter to reject Multipath TCP connections received on its Internet-facing interfaces. Only Multipath PTCP connections received on the customer-facing interfaces of a Converter will be accepted.
IANA is requested to assign a TCP port number (TBA) for the Converter Protocol from the "Service Name and Transport Protocol Port Number Registry" available at https://www.iana.org/assignments/service-names-port-numbers/service-names-port-numbers.xhtml.
IANA is requested to create a new "The Converter Protocol (Convert)
Parameters" registry.
The following subsections detail new registries within "The Converter
Protocol (Convert) Parameters" registry.
IANA is requested to create the "Convert versions" sub-registry. New
values are assigned via Standards Action.
The initial values to be assigned at the creation of the registry are
as follows:
+---------+--------------------------------------+-------------+ | Version | Description | Reference | +---------+--------------------------------------+-------------+ | 0 | Reserved by this document | [This-RFC] | | 1 | Assigned by this document | [This-RFC] | +---------+--------------------------------------+-------------+
IANA is requested to create the "Convert TLVs" sub-registry. The procedure for assigning values from this registry is as follows:
The initial values to be assigned at the creation of the registry are
as follows:
+---------+--------------------------------------+-------------+ | Code | Name | Reference | +---------+--------------------------------------+-------------+ | 0 | Reserved | [This-RFC] | | 1 | Bootstrap TLV | [This-RFC] | | 10 | Connect TLV | [This-RFC] | | 20 | Extended TCP Header TLV | [This-RFC] | | 22 | Supported TCP Extension Services TLV | [This-RFC] | | 30 | Error TLV | [This-RFC] | +---------+--------------------------------------+-------------+
IANA is requested to create the "Convert Errors" sub-registry. Codes in this registry are assigned as a function of the error type. Four types are defined; the following ranges are reserved for each of these types:
The procedure for assigning values from this sub-registry is as follows:
The initial values to be assigned at the creation of the registry are
as follows:
+-------+------+-----------------------------------+-----------+ | Error | Hex | Description | Reference | +-------+------+-----------------------------------+-----------+ | 0 | 0x00 | Unsupported Version | [This-RFC]| | 1 | 0x01 | Malformed Message | [This-RFC]| | 2 | 0x02 | Unsupported Message | [This-RFC]| | 32 | 0x20 | Not Authorized | [This-RFC]| | 33 | 0x21 | Unsupported TCP Option | [This-RFC]| | 64 | 0x40 | Resource Exceeded | [This-RFC]| | 65 | 0x41 | Network Failure | [This-RFC]| | 96 | 0x60 | Connection Reset | [This-RFC]| | 97 | 0x61 | Destination Unreachable | [This-RFC]| +-------+------+-----------------------------------+-----------+
Figure 19: The Convert Error Codes
Although they could disagree with the contents of the document, we would like to thank Joe Touch and Juliusz Chroboczek whose comments on the MPTCP mailing list have forced us to reconsider the design of the solution several times.
We would like to thank Raphael Bauduin, Stefano Secci, and Benjamin Hesmans for their help in preparing this document. Sri Gundavelli and Nandini Ganesh provided valuable feedback about the handling of TFO and the error codes. Thanks to them.
This document builds upon earlier documents that proposed various forms of Multipath TCP proxies [I-D.boucadair-mptcp-plain-mode], [I-D.peirens-mptcp-transparent] and [HotMiddlebox13b].
From [I-D.boucadair-mptcp-plain-mode]:
Many thanks to Chi Dung Phung, Mingui Zhang, Rao Shoaib, Yoshifumi Nishida, and Christoph Paasch for their valuable comments.
Thanks to Ian Farrer, Mikael Abrahamsson, Alan Ford, Dan Wing, and Sri Gundavelli for the fruitful discussions in IETF#95 (Buenos Aires).
Special thanks to Pierrick Seite, Yannick Le Goff, Fred Klamm, and Xavier Grall for their inputs.
Thanks also to Olaf Schleusing, Martin Gysi, Thomas Zasowski, Andreas Burkhard, Silka Simmen, Sandro Berger, Michael Melloul, Jean-Yves Flahaut, Adrien Desportes, Gregory Detal, Benjamin David, Arun Srinivasan, and Raghavendra Mallya for the discussion.
As noted above, this document builds on two previous documents.
The authors of [I-D.boucadair-mptcp-plain-mode] were: - Mohamed Boucadair - Christian Jacquenet - Olivier Bonaventure - Denis Behaghel - Stefano Secci - Wim Henderickx - Robert Skog - Suresh Vinapamula - SungHoon Seo - Wouter Cloetens - Ullrich Meyer - Luis M. Contreras - Bart Peirens
The authors of [I-D.peirens-mptcp-transparent] were: - Bart Peirens - Gregory Detal - Sebastien Barre - Olivier Bonaventure
This section to be removed before publication.
| [RFC0793] | Postel, J., "Transmission Control Protocol", STD 7, RFC 793, DOI 10.17487/RFC0793, September 1981. |
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
| [RFC4291] | Hinden, R. and S. Deering, "IP Version 6 Addressing Architecture", RFC 4291, DOI 10.17487/RFC4291, February 2006. |
| [RFC4727] | Fenner, B., "Experimental Values In IPv4, IPv6, ICMPv4, ICMPv6, UDP, and TCP Headers", RFC 4727, DOI 10.17487/RFC4727, November 2006. |
| [RFC4987] | Eddy, W., "TCP SYN Flooding Attacks and Common Mitigations", RFC 4987, DOI 10.17487/RFC4987, August 2007. |
| [RFC5482] | Eggert, L. and F. Gont, "TCP User Timeout Option", RFC 5482, DOI 10.17487/RFC5482, March 2009. |
| [RFC5925] | Touch, J., Mankin, A. and R. Bonica, "The TCP Authentication Option", RFC 5925, DOI 10.17487/RFC5925, June 2010. |
| [RFC6824] | Ford, A., Raiciu, C., Handley, M. and O. Bonaventure, "TCP Extensions for Multipath Operation with Multiple Addresses", RFC 6824, DOI 10.17487/RFC6824, January 2013. |
| [RFC7413] | Cheng, Y., Chu, J., Radhakrishnan, S. and A. Jain, "TCP Fast Open", RFC 7413, DOI 10.17487/RFC7413, December 2014. |
| [RFC8126] | Cotton, M., Leiba, B. and T. Narten, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 8126, DOI 10.17487/RFC8126, June 2017. |
| [RFC8174] | Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, May 2017. |
The description above is a simplified description of the Converter protocol. At a first glance, the proposed solution could seem similar to the SOCKS v5 protocol [RFC1928]. This protocol is used to proxy TCP connections. The Client creates a connection to a SOCKS proxy, exchanges authentication information and indicates the destination address and port of the final server. At this point, the SOCKS proxy creates a connection towards the final server and relays all data between the two proxied connections. The operation of an implementation based on SOCKSv5 is illustrated in Figure 20.
Client SOCKS Proxy Server
-------------------->
SYN
<--------------------
SYN+ACK
-------------------->
ACK
-------------------->
Version=5, Auth Methods
<--------------------
Method
-------------------->
Auth Request (unless "No auth" method negotiated)
<--------------------
Auth Response
-------------------->
Connect Server:Port -------------------->
SYN
<--------------------
SYN+ACK
<--------------------
Succeeded
-------------------->
Data1
-------------------->
Data1
<--------------------
Data2
<--------------------
Data2
Figure 20: Establishment of a TCP connection through a SOCKS proxy without authentication
The Converter protocol also relays data between an upstream and a downstream connection, but there are important differences with SOCKSv5.
A first difference is that the Converter protocol leverages the TFO option [RFC7413] to exchange all control information during the three-way handshake. This reduces the connection establishment delay compared to SOCKS that requires two or more round-trip-times before the establishment of the downstream connection towards the final destination. In today's Internet, latency is a important metric and various protocols have been tuned to reduce their latency [I-D.arkko-arch-low-latency]. A recently proposed extension to SOCKS also leverages the TFO option [I-D.olteanu-intarea-socks-6].
A second difference is that the Converter protocol explicitly takes the TCP extensions into account. By using the Converter protocol, the Client can learn whether a given TCP extension is supported by the destination Server. This enables the Client to bypass the Transport Converter when the destination supports the required TCP extension. Neither SOCKS v5 [RFC1928] nor the proposed SOCKS v6 [I-D.olteanu-intarea-socks-6] provide such a feature.
A third difference is that a Transport Converter will only accept the connection initiated by the Client provided that the downstream connection is accepted by the Server. If the Server refuses the connection establishment attempt from the Transport Converter, then the upstream connection from the Client is rejected as well. This feature is important for applications that check the availability of a Server or use the time to connect as a hint on the selection of a Server [RFC6555].