Internet Engineering Task Force (IETF) J. Touch Request for Comments: 6994 USC/ISI Category: Standards Track August 2013 ISSN: 2070-1721 Shared Use of Experimental TCP Options Abstract This document describes how the experimental TCP option codepoints can concurrently support multiple TCP extensions, even within the same connection, using a new IANA TCP experiment identifier. This approach is robust to experiments that are not registered and to those that do not use this sharing mechanism. It is recommended for all new TCP options that use these codepoints. Status of This Memo This is an Internet Standards Track document. This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 5741. Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc6994. Copyright Notice Copyright (c) 2013 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Table of Contents 1. Introduction ....................................................2 2. Conventions Used in This Document ...............................3 3. TCP Experimental Option Structure ...............................4 3.1. Selecting an ExID ..........................................5 3.2. Impact on TCP Option Processing ............................6 4. Reducing the Impact of False Positives ..........................7 5. Migration to Assigned Options ...................................7 6. Rationale .......................................................8 7. Security Considerations .........................................9 8. IANA Considerations .............................................9 9. References .....................................................10 9.1. Normative References ......................................10 9.2. Informative References ....................................10 10. Acknowledgments ...............................................11 1. Introduction TCP includes options to enable new protocol capabilities that can be activated only where needed and supported [RFC793]. The space for identifying such options is small -- 256 values, of which 30 are assigned at the time of this document's publication [IANA]. Two of these codepoints (253, 254) are allocated to support experiments [RFC4727]. These values are intended for testing purposes or for use when an assigned codepoint is either not warranted or available, e.g., based on the maturity status of the defined capability (i.e., Experimental or Informational, rather than Standards Track). Here, the term "experimental TCP options" refers to options that use the TCP experimental option codepoints [RFC4727]. Such experiments can be described in an RFC of any status (e.g., Experimental, Informational, etc.) and are intended to be used in controlled environments and are allowed in public deployments (when not enabled as default [RFC3692]). Nothing prohibits the deployment of multiple experiments in the same environment -- controlled or public. Further, some protocols are specified in Experimental or Informational RFCs, which either include parameters or design choices not yet understood or which might not be widely deployed [RFC2026]. Typically, these TCP options are not eligible to receive assigned codepoints [RFC2780], so they need a way to share their use of the experimental codepoints. There is currently no mechanism to support shared use of the TCP experimental option codepoints, either by different experiments on different connections or for more than two experimental options in the same connection. Experimental options 253 and 254 are already deployed in operational code to support an early version of TCP
authentication. Option 253 is also documented for the experimental TCP Cookie Transaction option [RFC6013]. This shared use results in collisions in which a single codepoint can appear multiple times in a single TCP segment and for which each use is ambiguous. Other codepoints have been used without assignment (known as "squatting"), notably 31-32 (TCP cookie transactions, as originally distributed and in its API doc) and 76-78 (tcpcrypt) [Bi11] [Si11]. Commercial products reportedly also use unassigned options 33, 69-70, and 76-78. Even though these uses are unauthorized, they currently impact legitimate assignees. Both such misuses (squatting on both experimental and assigned codepoints) are expected to continue, but there are several approaches that can alleviate the impact on cooperating protocol designers. One proposal relaxes the requirements for assignment of TCP options, allowing them to be assigned more readily for protocols that have not been standardized through the IETF process [RFC5226]. Another proposal assigns a larger pool to the TCP experiment option codepoints and manages their sharing through IANA coordination [Ed11]. The approach proposed in this document does not require additional TCP option codepoints and is robust to those who choose either not to support it or not to register their experiments. The solution adds a field to the structure of the experimental TCP option. This field is populated with an "experiment identifier" (ExID) defined as part of a specific option experiment. The ExID helps reduce the probability of a collision of independent experimental uses of the same option codepoint, both for those who follow this document (using registered ExIDs) and those who do not (squatters who either ignore this extension or do not register their ExIDs). The solution proposed in this document is recommended for all new protocols that use TCP experimental option codepoints. The techniques described here may also enable shared use of other experimental codepoints, but that issue is out of scope for this document. 2. Conventions Used in This Document The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119]. In this document, these words will appear with that interpretation only when in ALL CAPS. Lowercase uses of these words are not to be interpreted as carrying RFC 2119 significance.
In this document, the characters ">>" preceding an indented line(s) indicates a compliance requirement statement using the key words listed above. This convention aids reviewers in quickly identifying or finding the explicit compliance requirements of this RFC. 3. TCP Experimental Option Structure TCP options have the current common structure [RFC793], in which the first byte is the codepoint (Kind) and the second byte is the length of the option in bytes (Length): 0 1 2 3 01234567 89012345 67890123 45678901 +--------+--------+--------+--------+ | Kind | Length | ... | +--------+--------+--------+--------+ | ... +-------- Figure 1. TCP Option Structure [RFC793] This document extends the option structure for experimental codepoints (253, 254) with an experiment identifier (ExID), which is either 2 or 4 bytes in length. The ExID is used to differentiate experiments and is the first field after Kind and Length, as follows: 0 1 2 3 01234567 89012345 67890123 45678901 +--------+--------+--------+--------+ | Kind | Length | ExID | +--------+--------+--------+--------+ | option contents... +--------+--------+--------+--- Figure 2. TCP Experimental Option with a 16-bit ExID 0 1 2 3 01234567 89012345 67890123 45678901 +--------+--------+--------+--------+ | Kind | Length | ExID | +--------+--------+--------+--------+ | ExID (con't) | option contents... +--------+--------+--------+--- Figure 3. TCP Experimental Option with a 32-bit ExID
This mechanism is encouraged for all TCP options that are not yet eligible for assigned codepoints: >> Protocols requiring new TCP option codepoints that are not eligible for assigned values SHOULD use the existing TCP experimental option codepoints (253, 254) with ExIDs as described in this document. This mechanism is encouraged for all TCP options using the current experimental codepoints in controlled environments: >> All protocols using the TCP experimental option codepoints (253, 254), even those deployed in controlled environments, SHOULD use ExIDs as described in this document. This mechanism is required for all TCP options using the current experimental codepoints that are publicly deployed, whether enabled by default or not: >> All protocols using the TCP experimental option codepoints (253, 254) that are deployed outside controlled environments, such as in the public Internet, MUST use ExIDs as described in this document. Once a TCP option uses the mechanism in this document, registration of the ExID with IANA is required: >> All protocols using ExIDs as described in this document MUST register those ExIDs with IANA. Applicants register their desired ExID by contacting IANA [IANA]. 3.1. Selecting an ExID ExIDs are selected at design time, when the protocol designer first implements or specifies the experimental option. ExIDs can be either 16 bits or 32 bits. In both cases, the value is stored in the header in network-standard (big-endian) byte order. ExIDs combine properties of IANA registered codepoints with "magic numbers". >> All ExIDs MUST be either 16 bits or 32 bits long. Use of the ExID, whether 16 bit or 32 bit, helps reduce the probability of a false positive collision with those who either do not register their experiment or who do not implement this mechanism.
In order to conserve TCP option space, either for use within a specific option or to be available for other options: >> Options implementing the mechanism of this document SHOULD use 16-bit ExIDs, except where explicitly motivating the need for 32-bit ExIDs, e.g., to avoid false positives or maintain alignment with an expected future assigned codepoint. ExIDs are registered with IANA using "first come, first served" (FCFS) priority based on the first two bytes. Those two bytes are thus sufficient to interpret which experimental option is contained in the option field. >> All ExIDs MUST be unique based on their first 16 bits. The second two bytes serve as a "magic number". A magic number is a self-selected codepoint whose primary value is its unlikely collision with values selected by others. Magic numbers are used in other protocols, e.g., bootstrap protocol (BOOTP) [RFC951] and DHCP [RFC2131]. Using the additional magic number bytes helps the option contents have the same byte alignment in the TCP header as they would have if (or when) a conventional (non-experiment) TCP option codepoint is assigned. Use of the same alignment reduces the potential for implementation errors, especially in using the same word-alignment padding, if the same software is later modified to use a conventional codepoint. Use of the longer, 32-bit ExID further decreases the probability of such a false positive compared to those using shorter, 16-bit ExIDs. Use of the ExID does consume TCP option space but enables concurrent use of the experimental codepoints and provides protection against false positives, leaving less space for other options (including other experiments). Use of the longer, 32-bit ExID consumes more space, but provides more protection against false positives. 3.2. Impact on TCP Option Processing The ExID number is considered part of the TCP option, not the TCP option header. The presence of the ExID increases the effective option Length field by the size of the ExID. The presence of this ExID is thus transparent to implementations that do not support TCP options. During TCP processing, ExIDs in experimental options are matched against the ExIDs for each implemented protocol. The remainder of the option is specified by the particular experimental protocol.
>> Experimental options with ExIDs that do not match implemented protocols MUST be ignored. The ExID mechanism must be coordinated during connection establishment, just as with any TCP option. >> TCP ExID, if used in any TCP segment of a connection, MUST be present in TCP SYN segments of that connection. >> TCP experimental option ExIDs, if used in any TCP segment of a connection, SHOULD be used in all TCP segments of that connection in which any experimental option is present. Use of an ExID uses additional space in the TCP header and requires additional protocol processing by experimental protocols. Because these are experiments, neither consideration is a substantial impediment; a finalized protocol can avoid both issues with the assignment of a dedicated option codepoint later. 4. Reducing the Impact of False Positives False positives occur where the registered ExID of an experiment matches the value of an option that does not use ExIDs. Such collisions can cause an option to be interpreted by the incorrect processing routine. Use of checksums or signatures may help an experiment use the shorter ExID while reducing the corresponding increased potential for false positives. >> Experiments that are not robust to ExID false positives SHOULD implement other detection measures, such as checksums or minimal digital signatures over the experimental options they support. 5. Migration to Assigned Options Some experiments may transition away from being experimental and become eligible for an assigned TCP option codepoint. This document does not recommend a specific migration plan to transition from use of the experimental TCP options/ExIDs to use of an assigned codepoint. However, once an assigned codepoint is allocated, use of an ExID represents unnecessary overhead. As a result: >> Once a TCP option codepoint is assigned to a protocol, that protocol SHOULD NOT continue to use an ExID as part of that assigned codepoint.
This document does not recommend whether or how an implementation of an assigned codepoint can be backward compatible with use of the experimental codepoint/ExID. However, some implementers may be tempted to include both the experimental and assigned codepoint in the same segment, e.g., in a SYN to support backward compatibility during connection establishment. This is a poor use of limited resources; so, to ensure conservation of the TCP option space: >> A TCP segment MUST NOT contain both an assigned TCP option codepoint and a TCP experimental option codepoint for the same protocol. Instead, a TCP that intends backward compatibility might send multiple SYNs with alternates of the same option and discard all but the most desired successful connection. Although this approach may resolve more slowly or require additional effort at the endpoints, it is preferable to excessively consuming TCP option space. 6. Rationale The ExIDs described in this document combine properties of IANA FCFS-registered values with magic numbers. Although IANA FCFS registries are common, so too are those who either fail to register or who 'squat' by deliberately using codepoints that are assigned to others. The approach in this document is intended to recognize this reality and be more robust to its consequences than would be a conventional IANA FCFS registry. Existing ID spaces were considered as ExIDs in the development of this mechanism, including IEEE Organizationally Unique Identifier (OUI) and IANA Private Enterprise Numbers (PENs) [IEEE802] [OUI] [RFC1155]. OUIs are 24-bit identifiers that are combined with 24 to 40 bits of privately assigned space to create identifiers that are commonly assigned to a unique piece of hardware. OUIs are already longer than the smaller ExID value, and obtaining an OUI is costly (currently $1,885.00 USD). An OUI could be obtained for each experiment, but this could be considered expensive. An OUI already assigned to an organization could be shared if extended (to support multiple experiments within an organization), but this would either require coordination within an organization or an IANA registry; the former is prohibitive, and the latter is more complicated than having IANA manage the entire space.
PENs were originally used in the Simple Network Management Protocol (SNMP) [RFC1157]. PENs are identifiers that can be obtained without cost from IANA [PEN]. Despite the current registry, the size of the PEN assignment space is currently undefined and has only recently been proposed (as 32 bits) [IANA-PEN]. PENs are currently assigned to organizations, and there is no current process for assigning them to individuals. Finally, if the PENs are 32 bits as expected, they would be larger than needed in many cases. 7. Security Considerations The mechanism described in this document is not intended to enhance, nor does it weaken the existing state of security for TCP option processing. 8. IANA Considerations IANA has created a "TCP Experimental Option Experiment Identifiers (TCP ExIDs)" registry. The registry records both 16-bit and 32-bit ExIDs, as well as a reference (description, document pointer, or assignee name and e-mail contact) for each entry. ExIDs are registered for use with both of the TCP experimental option codepoints, i.e., with TCP options with values of 253 and 254. Entries are assigned on a First Come, First Served (FCFS) basis [RFC5226]. The registry operates FCFS on the first two bytes of the ExID (in network-standard order) but records the entire ExID (in network-standard order). Some examples are: o 0x12340000 collides with a previous registration of 0x1234abcd o 0x5678 collides with a previous registration of 0x56780123 o 0xabcd1234 collides with a previous registration of 0xabcd IANA will advise applicants of duplicate entries to select an alternate value, as per typical FCFS processing. IANA will record known duplicate uses to assist the community in both debugging assigned uses as well as correcting unauthorized duplicate uses. IANA should impose no requirements on making a registration other than indicating the desired codepoint and providing a point of contact. A short description or acronym for the use is desired but should not be required.
9. References 9.1. Normative References [RFC793] Postel, J., "Transmission Control Protocol", STD 7, RFC 793, September 1981. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC4727] Fenner, B., "Experimental Values In IPv4, IPv6, ICMPv4, ICMPv6, UDP, and TCP Headers", RFC 4727, November 2006. [RFC5226] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 5226, May 2008. 9.2. Informative References [Bi11] Bittau, A., Boneh, D., Hamburg, M., Handley, M., Mazieres, D., and Q. Slack, "Cryptographic protection of TCP Streams", Work in Progress, September 2012. [Ed11] Eddy, W., "Additional TCP Experimental-Use Options", Work in Progress, August 2011. [IANA] IANA, <http://www.iana.org/>. [IANA-PEN] Liang, P. and A. Melnikov, "Private Enterprise Number (PEN) Practices and Internet Assigned Numbers: Authority (IANA) Considerations for Registration Procedures", Work in Progress, June 2012. [IEEE802] IEEE, "IEEE Standard for Local and Metropolitan Area Networks: Overview and Architecture", IEEE 802-2001, 8 March 2002. [OUI] IEEE, "Organizationally Unique Identifier (OUI) or 'Company_ID'", <http://standards.ieee.org/develop/regauth/oui/>. [PEN] IANA, "Private Enterprise Numbers", <http://www.iana.org/assignments/enterprise-numbers>. [RFC951] Croft, W. and J. Gilmore, "Bootstrap Protocol", RFC 951, September 1985.
[RFC1155] Rose, M. and K. McCloghrie, "Structure and Identification of Management Information for TCP/IP-Based Internets", STD 16, RFC 1155, May 1990. [RFC1157] Case, J., Fedor, M., Schoffstall, M., and J. Davin, "Simple Network Management Protocol (SNMP)", RFC 1157, May 1990. [RFC2026] Bradner, S., "The Internet Standards Process -- Revision 3", BCP 9, RFC 2026, October 1996. [RFC2131] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131, March 1997. [RFC2780] Bradner, S. and V. Paxson, "IANA Allocation Guidelines For Values In the Internet Protocol and Related Headers", BCP 37, RFC 2780, March 2000. [RFC3692] Narten, T., "Assigning Experimental and Testing Numbers Considered Useful", BCP 82, RFC 3692, January 2004. [RFC6013] Simpson, W., "TCP Cookie Transactions (TCPCT)", RFC 6013, January 2011. [Si11] Simpson, W., "TCP Cookie Transactions (TCPCT) Sockets Application Program Interface (API)", Work in Progress, April 2011. 10. Acknowledgments This document was motivated by discussions on the IETF TCPM mailing list and by Wes Eddy's proposal [Ed11]. Yoshifumi Nishida, Pasi Sarolathi, and Michael Scharf provided detailed feedback. This document was originally prepared using 2-Word-v2.0.template.dot. Author's Address Joe Touch USC/ISI 4676 Admiralty Way Marina del Rey, CA 90292-6695 U.S.A. Phone: +1 (310) 448-9151 EMail: email@example.com