TAPS M. Welzl
Internet-Draft S. Gjessing
Intended status: Informational University of Oslo
Expires: December 7, 2018 June 5, 2018

A Minimal Set of Transport Services for End Systems
draft-ietf-taps-minset-04

Abstract

This draft recommends a minimal set of Transport Services offered by end systems, and gives guidance on choosing among the available mechanisms and protocols. It is based on the set of transport features in RFC 8303.

Status of This Memo

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 December 7, 2018.

Copyright Notice

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.

Table of Contents

1. Introduction

The task of a transport system is to offer transport services to its applications, i.e. the applications running on top of the transport system. Ideally, it does so without statically binding applications to particular transport protocols. Currently, the set of transport services that most applications use is based on TCP and UDP (and protocols that are layered on top of them); this limits the ability for the network stack to make use of features of other transport protocols. For example, if a protocol supports out-of-order message delivery but applications always assume that the network provides an ordered bytestream, then the network stack can not immediately deliver a message that arrives out-of-order: doing so would break a fundamental assumption of the application. The net result is unnecessary head-of-line blocking delay.

By exposing the transport services of multiple transport protocols, a transport system can make it possible to use these services without having to statically bind an application to a specific transport protocol. The first step towards the design of such a system was taken by [RFC8095], which surveys a large number of transports, and [RFC8303] as well as [RFC8304], which identify the specific transport features that are exposed to applications by the protocols TCP, MPTCP, UDP(-Lite) and SCTP as well as the LEDBAT congestion control mechanism. This memo is based on these documents and follows the same terminology (also listed below). Because the considered transport protocols conjointly cover a wide range of transport features, there is reason to hope that the resulting set (and the reasoning that led to it) will also apply to many aspects of other transport protocols.

The number of transport features of current IETF transports is large, and exposing all of them has a number of disadvantages: generally, the more functionality is exposed, the less freedom a transport system has to automate usage of the various functions of its available set of transport protocols. Some functions only exist in one particular protocol, and if an application used them, this would statically tie the application to this protocol, limiting the flexibility of the transport system. Also, if the number of exposed features is exceedingly large, a transport system might become very difficult to use for an application programmer. Taking [RFC8303] as a basis, this document therefore develops a minimal set of transport features, removing the ones that could get in the way of transport flexibility but keeping the ones that must be retained for applications to benefit from useful transport functionality.

Applications use a wide variety of APIs today. The transport features in the minimal set in this document must be reflected in *all* network APIs in order for the underlying functionality to become usable everywhere. For example, it does not help an application that talks to a library which offers its own communication interface if only the underlying Berkeley Sockets API is extended to offer "unordered message delivery", but the library only exposes an ordered bytestream. Both the Berkeley Sockets API and the library would have to expose the "unordered message delivery" transport feature (alternatively, there may be ways for certain types of libraries to use this transport feature without exposing it, based on knowledge about the applications -- but this is not the general case). In most situations, in the interest of being as flexible and efficient as possible, the best choice will be for a library to expose at least all of the transport features that are recommended as a "minimal set" here.

This "minimal set" can be implemented one-sided over TCP (or UDP, if certain limitations are put in place). This means that a sender-side transport system can talk to a standard TCP (or UDP) receiver, and a receiver-side transport system can talk to a standard TCP (or UDP) sender.

2. Terminology

Transport Feature:
a specific end-to-end feature that the transport layer provides to an application. Examples include confidentiality, reliable delivery, ordered delivery, message-versus-stream orientation, etc.
Transport Service:
a set of Transport Features, without an association to any given framing protocol, which provides a complete service to an application.
Transport Protocol:
an implementation that provides one or more different transport services using a specific framing and header format on the wire.
Transport Service Instance:
an arrangement of transport protocols with a selected set of features and configuration parameters that implements a single transport service, e.g., a protocol stack (RTP over UDP).
Application:
an entity that uses the transport layer for end-to-end delivery data across the network (this may also be an upper layer protocol or tunnel encapsulation).
Application-specific knowledge:
knowledge that only applications have.
Endpoint:
an entity that communicates with one or more other endpoints using a transport protocol.
Connection:
shared state of two or more endpoints that persists across messages that are transmitted between these endpoints.
Socket:
the combination of a destination IP address and a destination port number.

Moreover, throughout the document, the protocol name "UDP(-Lite)" is used when discussing transport features that are equivalent for UDP and UDP-Lite; similarly, the protocol name "TCP" refers to both TCP and MPTCP.

3. The Minimal Set of Transport Features

Based on the categorization, reduction, and discussion in Appendix A, this section describes a minimal set of transport features that end systems should offer. The described transport system can be implemented over TCP. Elements of the system that are not marked with "!UDP" can also be implemented over UDP.

As in Appendix A, Appendix A.2 and [RFC8303], we categorize the minimal set of transport features as 1) CONNECTION related (ESTABLISHMENT, AVAILABILITY, MAINTENANCE, TERMINATION) and 2) DATA Transfer related (Sending Data, Receiving Data, Errors). Here, the focus is on connections that the transport system offers as an abstraction to the application, as opposed to connections of transport protocols that the transport system uses.

3.1. ESTABLISHMENT, AVAILABILITY and TERMINATION

A connection must first be "created" to allow for some initial configuration to be carried out before the transport system can actively or passively establish communication with a remote endpoint. All configuration parameters in Section 3.2 can be used initially, although some of them may only take effect when a connection has been established with a chosen transport protocol. Configuring a connection early helps a transport system make the right decisions. For example, grouping information can influence the transport system to implement a connection as a stream of a multi-streaming protocol's existing association or not. 


- Will it ever be necessary to offer any of the following?
  *  Reliably transfer data
  *  Notify the peer of closing/aborting
  *  Preserve data ordering

  Yes: SCTP or TCP can be used.
  - Is any of the following useful to the application?
    * Choosing a scheduler to operate between connections
      in a group, with the possibility to configure a priority
      or weight per connection
    * Configurable message reliability
    * Unordered message delivery
    * Request not to delay the acknowledgement (SACK) of a message

    Yes: SCTP is preferred.
    No:
    - Is any of the following useful to the application?
      * Hand over a message to reliably transfer (possibly
        multiple times) before connection establishment
      * Suggest timeout to the peer
      * Notification of Excessive Retransmissions (early
        warning below abortion threshold)
      * Notification of ICMP error message arrival
      
      Yes: TCP is preferred.
      No: SCTP and TCP are equally preferable.
      

  No: all protocols can be used.
  - Is any of the following useful to the application?
    *  Specify checksum coverage used by the sender
    *  Specify minimum checksum coverage required by receiver

    Yes: UDP-Lite is preferred.
    No: UDP is preferred.

For ungrouped connections, early configuration is necessary because it allows the transport system to know which protocols it should try to use. In particular, a transport system that only makes a one-time choice for a particular protocol must know early about strict requirements that must be kept, or it can end up in a deadlock situation (e.g., having chosen UDP and later be asked to support reliable transfer). As an example description of how to correctly handle these cases, we provide the following decision tree (this is derived from Appendix A.2.1 excluding authentication, as explained in Section 7):

Note that this decision tree is not optimal for all cases. For example, if an application wants to use "Specify checksum coverage used by the sender", which is only offered by UDP-Lite, and "Configure priority or weight for a scheduler", which is only offered by SCTP, the above decision tree will always choose UDP-Lite, making it impossible to use SCTP's schedulers with priorities between grouped connections. The transport system must know which choice is more important for the application in order to make the best decision. We caution implementers to be aware of the full set of trade-offs, for which we recommend consulting the list in Appendix A.2.1 when deciding how to initialize a connection.

To summarize, the following parameters serve as input for the transport system to help it choose and configure a suitable protocol:

Once a connection is created, it can be queried for the maximum amount of data that an application can possibly expect to have reliably transmitted before or during transport connection establishment (with zero being a possible answer) (see Section 3.2.1). An application can also give the connection a message for reliable transmission before or during connection establishment (!UDP); the transport system will then try to transmit it as early as possible. An application can facilitate sending a message particularly early by marking it as "idempotent" (see Section 3.3.1); in this case, the receiving application must be prepared to potentially receive multiple copies of the message (because idempotent messages are reliably transferred, asking for idempotence is not necessary for systems that support UDP).

After creation, a transport system can actively establish communication with a peer, or it can passively listen for incoming connection requests. Note that active establishment may or may not trigger a notification on the listening side. It is possible that the first notification on the listening side is the arrival of the first data that the active side sends (a receiver-side transport system could handle this by continuing to block a "Listen" call, immediately followed by issuing "Receive", for example; callback-based implementations could simply skip the equivalent of "Listen"). This also means that the active opening side is assumed to be the first side sending data.

A transport system can actively close a connection, i.e. terminate it after reliably delivering all remaining data to the peer (if reliable data delivery was requested earlier (!UDP)), in which case the peer is notified that the connection is closed. Alternatively, a connection can be aborted without delivering outstanding data to the peer. In case reliable or partially reliable data delivery was requested earlier (!UDP), the peer is notified that the connection is aborted. A timeout can be configured to abort a connection when data could not be delivered for too long (!UDP); however, timeout-based abortion does not notify the peer application that the connection has been aborted. Because half-closed connections are not supported, when a host implementing a transport system receives a notification that the peer is closing or aborting the connection (!UDP), its peer may not be able to read outstanding data. This means that unacknowledged data residing a transport system's send buffer may have to be dropped from that buffer upon arrival of a "close" or "abort" notification from the peer.

3.2. MAINTENANCE

A transport system must offer means to group connections, but it cannot guarantee truly grouping them using the transport protocols that it uses (e.g., it cannot be guaranteed that connections become multiplexed as streams on a single SCTP association when SCTP may not be available). The transport system must therefore ensure that group- versus non-group-configurations are handled correctly in some way (e.g., by applying the configuration to all grouped connections even when they are not multiplexed, or informing the application about grouping success or failure).

As a general rule, any configuration described below should be carried out as early as possible to aid the transport system's decision making.

3.2.1. Connection groups

The following transport features and notifications (some directly from Appendix A.2, some new or changed, based on the discussion in Appendix A.3) automatically apply to all grouped connections:

(!UDP) Configure a timeout: this can be done with the following parameters:

Configure urgency: this can be done with the following parameters:

Following Appendix A.3.7, these properties can be queried:



In addition to the already mentioned closing / aborting notifications and possible send errors, the following notifications can occur:

3.2.2. Individual connections

Configure priority or weight for a scheduler, as described in [RFC8260].

Configure checksum usage: this can be done with the following parameters, but there is no guarantee that any checksum limitations will indeed be enforced (the default behavior is "full coverage, checksum enabled"):

3.3. DATA Transfer

3.3.1. Sending Data

When sending a message, no guarantees are given about the preservation of message boundaries to the peer; if message boundaries are needed, the receiving application at the peer must know about them beforehand (or the transport system cannot use TCP). Note that an application should already be able to hand over data before the transport system establishes a connection with a chosen transport protocol. Regarding the message that is being handed over, the following parameters can be used:

An application can be notified of a failure to send a specific message. There is no guarantee of such notifications, i.e. send failures can also silently occur.

3.3.2. Receiving Data

A receiving application obtains an "Application-Framed Bytestream" (AFra-Bytestream); this concept is further described in Appendix A.3.1). In line with TCP's receiver semantics, an AFra-Bytestream is just a stream of bytes to the receiver. If message boundaries were specified by the sender, a receiver-side transport system implementing only the minimum set of transport services defined here will still not inform the receiving application about them (this limitation is only needed for transport systems that are implemented to directly use TCP).

Different from TCP's semantics, if the sending application has allowed that messages are not fully reliably transferred, or delivered out of order, then such re-ordering or unreliability may be reflected per message in the arriving data. Messages will always stay intact - i.e. if an incomplete message is contained at the end of the arriving data block, this message is guaranteed to continue in the next arriving data block.

4. Conclusion

By decoupling applications from transport protocols, a transport system provides a different abstraction level than the Berkeley sockets interface. As with high- vs. low-level programming languages, a higher abstraction level allows more freedom for automation below the interface, yet it takes some control away from the application programmer. This is the design trade-off that a transport system developer is facing, and this document provides guidance on the design of this abstraction level. Some transport features are currently rarely offered by APIs, yet they must be offered or they can never be used ("functional" transport features). Other transport features are offered by the APIs of the protocols covered here, but not exposing them in an API would allow for more freedom to automate protocol usage in a transport system. The minimal set presented in this document is an effort to find a middle ground that can be recommended for transport systems to implement, on the basis of the transport features discussed in [RFC8303].

5. Acknowledgements

The authors would like to thank all the participants of the TAPS Working Group and the NEAT and MAMI research projects for valuable input to this document. We especially thank Michael Tuexen for help with connection connection establishment/teardown and Gorry Fairhurst for his suggestions regarding fragmentation and packet sizes. This work has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 644334 (NEAT).

6. IANA Considerations

XX RFC ED - PLEASE REMOVE THIS SECTION XXX

This memo includes no request to IANA.

7. Security Considerations

Authentication, confidentiality protection, and integrity protection are identified as transport features by [RFC8095]. As currently deployed in the Internet, these features are generally provided by a protocol or layer on top of the transport protocol; no current full-featured standards-track transport protocol provides all of these transport features on its own. Therefore, these transport features are not considered in this document, with the exception of native authentication capabilities of TCP and SCTP for which the security considerations in [RFC5925] and [RFC4895] apply. The minimum security requirements for a transport system are discussed in a separate document [I-D.ietf-taps-transport-security].

8. References

8.1. Normative References

[RFC8303] Welzl, M., Tuexen, M. and N. Khademi, "On the Usage of Transport Features Provided by IETF Transport Protocols", RFC 8303, DOI 10.17487/RFC8303, February 2018.

8.2. Informative References

[COBS] Cheshire, S. and M. Baker, "Consistent Overhead Byte Stuffing", September 1997.
[I-D.ietf-taps-transport-security] Pauly, T., Perkins, C., Rose, K. and C. Wood, "A Survey of Transport Security Protocols", Internet-Draft draft-ietf-taps-transport-security-01, May 2018.
[I-D.ietf-tsvwg-rtcweb-qos] Jones, P., Dhesikan, S., Jennings, C. and D. Druta, "DSCP Packet Markings for WebRTC QoS", Internet-Draft draft-ietf-tsvwg-rtcweb-qos-18, August 2016.
[LBE-draft] Bless, R., "A Lower Effort Per-Hop Behavior (LE PHB)", Internet-draft draft-tsvwg-le-phb-03, February 2018.
[RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, RFC 2914, DOI 10.17487/RFC2914, September 2000.
[RFC3758] Stewart, R., Ramalho, M., Xie, Q., Tuexen, M. and P. Conrad, "Stream Control Transmission Protocol (SCTP) Partial Reliability Extension", RFC 3758, DOI 10.17487/RFC3758, May 2004.
[RFC4895] Tuexen, M., Stewart, R., Lei, P. and E. Rescorla, "Authenticated Chunks for the Stream Control Transmission Protocol (SCTP)", RFC 4895, DOI 10.17487/RFC4895, August 2007.
[RFC4987] Eddy, W., "TCP SYN Flooding Attacks and Common Mitigations", RFC 4987, DOI 10.17487/RFC4987, August 2007.
[RFC5925] Touch, J., Mankin, A. and R. Bonica, "The TCP Authentication Option", RFC 5925, DOI 10.17487/RFC5925, June 2010.
[RFC7305] Lear, E., "Report from the IAB Workshop on Internet Technology Adoption and Transition (ITAT)", RFC 7305, DOI 10.17487/RFC7305, July 2014.
[RFC7413] Cheng, Y., Chu, J., Radhakrishnan, S. and A. Jain, "TCP Fast Open", RFC 7413, DOI 10.17487/RFC7413, December 2014.
[RFC7496] Tuexen, M., Seggelmann, R., Stewart, R. and S. Loreto, "Additional Policies for the Partially Reliable Stream Control Transmission Protocol Extension", RFC 7496, DOI 10.17487/RFC7496, April 2015.
[RFC8095] Fairhurst, G., Trammell, B. and M. Kuehlewind, "Services Provided by IETF Transport Protocols and Congestion Control Mechanisms", RFC 8095, DOI 10.17487/RFC8095, March 2017.
[RFC8260] Stewart, R., Tuexen, M., Loreto, S. and R. Seggelmann, "Stream Schedulers and User Message Interleaving for the Stream Control Transmission Protocol", RFC 8260, DOI 10.17487/RFC8260, November 2017.
[RFC8304] Fairhurst, G. and T. Jones, "Transport Features of the User Datagram Protocol (UDP) and Lightweight UDP (UDP-Lite)", RFC 8304, DOI 10.17487/RFC8304, February 2018.
[WWDC2015] Lakhera, P. and S. Cheshire, "Your App and Next Generation Networks", Apple Worldwide Developers Conference 2015, San Francisco, USA, June 2015.

Appendix A. Deriving the minimal set

We approach the construction of a minimal set of transport features in the following way: Section 3.

  1. Categorization: the superset of transport features from [RFC8303] is presented, and transport features are categorized for later reduction.
  2. Reduction: a shorter list of transport features is derived from the categorization in the first step. This removes all transport features that do not require application-specific knowledge or cannot be implemented with TCP or UDP.
  3. Discussion: the resulting list shows a number of peculiarities that are discussed, to provide a basis for constructing the minimal set.
  4. Construction: Based on the reduced set and the discussion of the transport features therein, a minimal set is constructed.

The first three steps as well as the underlying rationale for constructing the minimal set are described in this appendix. The minimal set itself is described in

A.1. Step 1: Categorization -- The Superset of Transport Features

Following [RFC8303], we divide the transport features into two main groups as follows:

  1. CONNECTION related transport features
    - ESTABLISHMENT
    - AVAILABILITY
    - MAINTENANCE
    - TERMINATION
  2. DATA Transfer related transport features
    - Sending Data
    - Receiving Data
    - Errors

We assume that applications have no specific requirements that need knowledge about the network, e.g. regarding the choice of network interface or the end-to-end path. Even with these assumptions, there are certain requirements that are strictly kept by transport protocols today, and these must also be kept by a transport system. Some of these requirements relate to transport features that we call "Functional".

Functional transport features provide functionality that cannot be used without the application knowing about them, or else they violate assumptions that might cause the application to fail. For example, ordered message delivery is a functional transport feature: it cannot be configured without the application knowing about it because the application's assumption could be that messages always arrive in order. Failure includes any change of the application behavior that is not performance oriented, e.g. security.

"Change DSCP" and "Disable Nagle algorithm" are examples of transport features that we call "Optimizing": if a transport system autonomously decides to enable or disable them, an application will not fail, but a transport system may be able to communicate more efficiently if the application is in control of this optimizing transport feature. These transport features require application-specific knowledge (e.g., about delay/bandwidth requirements or the length of future data blocks that are to be transmitted).

The transport features of IETF transport protocols that do not require application-specific knowledge and could therefore be transparently utilized by a transport system are called "Automatable".

Finally, some transport features are aggregated and/or slightly changed in the description below. These transport features are marked as "ADDED". The corresponding transport features are automatable, and they are listed immediately below the "ADDED" transport feature.

In this description, transport services are presented following the nomenclature "CATEGORY.[SUBCATEGORY].SERVICENAME.PROTOCOL", equivalent to "pass 2" in [RFC8303]. We also sketch how some of the transport features can be implemented by a transport system. For all transport features that are categorized as "functional" or "optimizing", and for which no matching TCP and/or UDP primitive exists in "pass 2" of [RFC8303], a brief discussion on how to implement them over TCP and/or UDP is included.

We designate some transport features as "automatable" on the basis of a broader decision that affects multiple transport features:

A.1.1. CONNECTION Related Transport Features

ESTABLISHMENT:

AVAILABILITY:

MAINTENANCE:

TERMINATION:

A.1.2. DATA Transfer Related Transport Features

A.1.2.1. Sending Data

A.1.2.2. Receiving Data

A.1.2.3. Errors

This section describes sending failures that are associated with a specific call to in the "Sending Data" category (Appendix A.1.2.1).

A.2. Step 2: Reduction -- The Reduced Set of Transport Features

By hiding automatable transport features from the application, a transport system can gain opportunities to automate the usage of network-related functionality. This can facilitate using the transport system for the application programmer and it allows for optimizations that may not be possible for an application. For instance, system-wide configurations regarding the usage of multiple interfaces can better be exploited if the choice of the interface is not entirely up to the application. Therefore, since they are not strictly necessary to expose in a transport system, we do not include automatable transport features in the reduced set of transport features. This leaves us with only the transport features that are either optimizing or functional.

A transport system should be able to communicate via TCP or UDP if alternative transport protocols are found not to work. For many transport features, this is possible -- often by simply not doing anything when a specific request is made. For some transport features, however, it was identified that direct usage of neither TCP nor UDP is possible: in these cases, even not doing anything would incur semantically incorrect behavior. Whenever an application would make use of one of these transport features, this would eliminate the possibility to use TCP or UDP. Thus, we only keep the functional and optimizing transport features for which an implementation over either TCP or UDP is possible in our reduced set.

In the following list, we precede a transport feature with "T:" if an implementation over TCP is possible, "U:" if an implementation over UDP is possible, and "TU:" if an implementation over either TCP or UDP is possible.

A.2.1. CONNECTION Related Transport Features

ESTABLISHMENT:

AVAILABILITY:

MAINTENANCE:

TERMINATION:

A.2.2. DATA Transfer Related Transport Features

A.2.2.1. Sending Data

A.2.2.2. Receiving Data

A.2.2.3. Errors

This section describes sending failures that are associated with a specific call to in the "Sending Data" category (Appendix A.1.2.1).

A.3. Step 3: Discussion

The reduced set in the previous section exhibits a number of peculiarities, which we will discuss in the following. This section focuses on TCP because, with the exception of one particular transport feature ("Receive a message" -- we will discuss this in Appendix A.3.1), the list shows that UDP is strictly a subset of TCP. We can first try to understand how to build a transport system that can run over TCP, and then narrow down the result further to allow that the system can always run over either TCP or UDP (which effectively means removing everything related to reliability, ordering, authentication and closing/aborting with a notification to the peer).

Note that, because the functional transport features of UDP are -- with the exception of "Receive a message" -- a subset of TCP, TCP can be used as a replacement for UDP whenever an application does not need message delimiting (e.g., because the application-layer protocol already does it). This has been recognized by many applications that already do this in practice, by trying to communicate with UDP at first, and falling back to TCP in case of a connection failure.

A.3.1. Sending Messages, Receiving Bytes

For implementing a transport system over TCP, there are several transport features related to sending, but only a single transport feature related to receiving: "Receive data (with no message delimiting)" (and, strangely, "information about partial message arrival"). Notably, the transport feature "Receive a message" is also the only non-automatable transport feature of UDP(-Lite) for which no implementation over TCP is possible.

To support these TCP receiver semantics, we define an "Application-Framed Bytestream" (AFra-Bytestream). AFra-Bytestreams allow senders to operate on messages while minimizing changes to the TCP socket API. In particular, nothing changes on the receiver side - data can be accepted via a normal TCP socket.

In an AFra-Bytestream, the sending application can optionally inform the transport about message boundaries and required properties per message (configurable order and reliability, or embedding a request not to delay the acknowledgement of a message). Whenever the sending application specifies per-message properties that relax the notion of reliable in-order delivery of bytes, it must assume that the receiving application is 1) able to determine message boundaries, provided that messages are always kept intact, and 2) able to accept these relaxed per-message properties. Any signaling of such information to the peer is up to an application-layer protocol and considered out of scope of this document.

For example, if an application requests to transfer fixed-size messages of 100 bytes with partial reliability, this needs the receiving application to be prepared to accept data in chunks of 100 bytes. If, then, some of these 100-byte messages are missing (e.g., if SCTP with Configurable Reliability is used), this is the expected application behavior. With TCP, no messages would be missing, but this is also correct for the application, and the possible retransmission delay is acceptable within the best effort service model [RFC7305]. Still, the receiving application would separate the byte stream into 100-byte chunks.

Note that this usage of messages does not require all messages to be equal in size. Many application protocols use some form of Type-Length-Value (TLV) encoding, e.g. by defining a header including length fields; another alternative is the use of byte stuffing methods such as COBS [COBS]. If an application needs message numbers, e.g. to restore the correct sequence of messages, these must also be encoded by the application itself, as the sequence number related transport features of SCTP are not provided by the "minimum set" (in the interest of enabling usage of TCP).

A.3.2. Stream Schedulers Without Streams

We have already stated that multi-streaming does not require application-specific knowledge. Potential benefits or disadvantages of, e.g., using two streams of an SCTP association versus using two separate SCTP associations or TCP connections are related to knowledge about the network and the particular transport protocol in use, not the application. However, the transport features "Choose a scheduler to operate between streams of an association" and "Configure priority or weight for a scheduler" operate on streams. Here, streams identify communication channels between which a scheduler operates, and they can be assigned a priority. Moreover, the transport features in the MAINTENANCE category all operate on assocations in case of SCTP, i.e. they apply to all streams in that assocation.

With only these semantics necessary to represent, the interface to a transport system becomes easier if we assume that connections may be a transport protocol's connection or association, but could also be a stream of an existing SCTP association, for example. We only need to allow for a way to define a possible grouping of connections. Then, all MAINTENANCE transport features can be said to operate on connection groups, not connections, and a scheduler operates on the connections within a group.

To be compatible with multiple transport protocols and uniformly allow access to both transport connections and streams of a multi-streaming protocol, the semantics of opening and closing need to be the most restrictive subset of all of the underlying options. For example, TCP's support of half-closed connections can be seen as a feature on top of the more restrictive "ABORT"; this feature cannot be supported because not all protocols used by a transport system (including streams of an association) support half-closed connections.

A.3.3. Early Data Transmission

There are two transport features related to transferring a message early: "Hand over a message to reliably transfer (possibly multiple times) before connection establishment", which relates to TCP Fast Open [RFC7413], and "Hand over a message to reliably transfer during connection establishment", which relates to SCTP's ability to transfer data together with the COOKIE-Echo chunk. Also without TCP Fast Open, TCP can transfer data during the handshake, together with the SYN packet -- however, the receiver of this data may not hand it over to the application until the handshake has completed. Also, different from TCP Fast Open, this data is not delimited as a message by TCP (thus, not visible as a ``message''). This functionality is commonly available in TCP and supported in several implementations, even though the TCP specification does not explain how to provide it to applications.

A transport system could differentiate between the cases of transmitting data "before" (possibly multiple times) or "during" the handshake. Alternatively, it could also assume that data that are handed over early will be transmitted as early as possible, and "before" the handshake would only be used for messages that are explicitly marked as "idempotent" (i.e., it would be acceptable to transfer them multiple times).

The amount of data that can successfully be transmitted before or during the handshake depends on various factors: the transport protocol, the use of header options, the choice of IPv4 and IPv6 and the Path MTU. A transport system should therefore allow a sending application to query the maximum amount of data it can possibly transmit before (or, if exposed, during) connection establishment.

A.3.4. Sender Running Dry

The transport feature "Notification that the stack has no more user data to send" relates to SCTP's "SENDER DRY" notification. Such notifications can, in principle, be used to avoid having an unnecessarily large send buffer, yet ensure that the transport sender always has data available when it has an opportunity to transmit it. This has been found to be very beneficial for some applications [WWDC2015]. However, "SENDER DRY" truly means that the entire send buffer (including both unsent and unacknowledged data) has emptied -- i.e., when it notifies the sender, it is already too late, the transport protocol already missed an opportunity to send data. Some modern TCP implementations now include the unspecified "TCP_NOTSENT_LOWAT" socket option that was proposed in [WWDC2015], which limits the amount of unsent data that TCP can keep in the socket buffer; this allows to specify at which buffer filling level the socket becomes writable, rather than waiting for the buffer to run empty.

SCTP allows to configure the sender-side buffer too: the automatable Transport Feature "Configure send buffer size" provides this functionality, but only for the complete buffer, which includes both unsent and unacknowledged data. SCTP does not allow to control these two sizes separately. It therefore makes sense for a transport system to allow for uniform access to "TCP_NOTSENT_LOWAT" as well as the "SENDER DRY" notification.

A.3.5. Capacity Profile

The transport features:

all relate to a QoS-like application need such as "low latency" or "scavenger". In the interest of flexibility of a transport system, they could therefore be offered in a uniform, more abstract way, where a transport system could e.g. decide by itself how to use combinations of LEDBAT-like congestion control and certain DSCP values, and an application would only specify a general "capacity profile" (a description of how it wants to use the available capacity). A need for "lowest possible latency at the expense of overhead" could then translate into automatically disabling the Nagle algorithm.

In some cases, the Nagle algorithm is best controlled directly by the application because it is not only related to a general profile but also to knowledge about the size of future messages. For fine-grain control over Nagle-like functionality, the "Request not to bundle messages" is available.

A.3.6. Security

Both TCP and SCTP offer authentication. TCP authenticates complete segments. SCTP allows to configure which of SCTP's chunk types must always be authenticated -- if this is exposed as such, it creates an undesirable dependency on the transport protocol. For compatibility with TCP, a transport system should only allow to configure complete transport layer packets, including headers, IP pseudo-header (if any) and payload.

Security is discussed in a separate document [I-D.ietf-taps-transport-security]. The minimal set presented in the present document therefore excludes all security related transport features: "Configure authentication", "Change authentication parameters", "Obtain authentication information" and and "Set Cookie life value" as well as "Specifying a key id to be used to authenticate a message".

A.3.7. Packet Size

UDP(-Lite) has a transport feature called "Specify DF field". This yields an error message in case of sending a message that exceeds the Path MTU, which is necessary for a UDP-based application to be able to implement Path MTU Discovery (a function that UDP-based applications must do by themselves). The "Get max. transport-message size that may be sent using a non-fragmented IP packet from the configured interface" transport feature yields an upper limit for the Path MTU (minus headers) and can therefore help to implement Path MTU Discovery more efficiently.

Appendix B. Revision information

XXX RFC-Ed please remove this section prior to publication.

-02: implementation suggestions added, discussion section added, terminology extended, DELETED category removed, various other fixes; list of Transport Features adjusted to -01 version of [RFC8303] except that MPTCP is not included.

-03: updated to be consistent with -02 version of [RFC8303].

-04: updated to be consistent with -03 version of [RFC8303]. Reorganized document, rewrote intro and conclusion, and made a first stab at creating a real "minimal set".

-05: updated to be consistent with -05 version of [RFC8303] (minor changes). Fixed a mistake regarding Cookie Life value. Exclusion of security related transport features (to be covered in a separate document). Reorganized the document (now begins with the minset, derivation is in the appendix). First stab at an abstract API for the minset.

draft-ietf-taps-minset-00: updated to be consistent with -08 version of [RFC8303] ("obtain message delivery number" was removed, as this has also been removed in [RFC8303] because it was a mistake in RFC4960. This led to the removal of two more transport features that were only designated as functional because they affected "obtain message delivery number"). Fall-back to UDP incorporated (this was requested at IETF-99); this also affected the transport feature "Choice between unordered (potentially faster) or ordered delivery of messages" because this is a boolean which is always true for one fall-back protocol, and always false for the other one. This was therefore now divided into two features, one for ordered, one for unordered delivery. The word "reliably" was added to the transport features "Hand over a message to reliably transfer (possibly multiple times) before connection establishment" and "Hand over a message to reliably transfer during connection establishment" to make it clearer why this is not supported by UDP. Clarified that the "minset abstract interface" is not proposing a specific API for all TAPS systems to implement, but it is just a way to describe the minimum set. Author order changed.

WG -01: "fall-back to" (TCP or UDP) replaced (mostly with "implementation over"). References to post-sockets removed (these were statments that assumed that post-sockets requires two-sided implementation). Replaced "flow" with "TAPS Connection" and "frame" with "message" to avoid introducing new terminology. Made sections 3 and 4 in line with the categorization that is already used in the appendix and [RFC8303], and changed style of section 4 to be even shorter and less interface-like. Updated reference draft-ietf-tsvwg-sctp-ndata to RFC8260.

WG -02: rephrased "the TAPS system" and "TAPS connection" etc. to more generally talk about transport after the intro (mostly replacing "TAPS system" with "transport system" and "TAPS connection" with "connection". Merged sections 3 and 4 to form a new section 3.

WG -03: updated sentence referencing [I-D.ietf-taps-transport-security] to say that "the minimum security requirements for a taps system are discussed in a separate security document", wrote "example" in the paragraph introducing the decision tree. Removed reference draft-grinnemo-taps-he-03 and the sentence that referred to it.

WG -04: addressed comments from Theresa Enghardt and Tommy Pauly. As part of that, removed "TAPS" as a term everywhere (abstract, intro, ..).

Authors' Addresses

Michael Welzl University of Oslo PO Box 1080 Blindern Oslo, N-0316 Norway Phone: +47 22 85 24 20 EMail: michawe@ifi.uio.no
Stein Gjessing University of Oslo PO Box 1080 Blindern Oslo, N-0316 Norway Phone: +47 22 85 24 44 EMail: steing@ifi.uio.no