Transport Area Report
Transport Area Working Groups:
Audio/Video Transport (avt)
The AVT working group met for two very full sessions. The group agreed that WG last call should be issued on the drafts for "Options for Repair of Streaming Media" and H.263+ payload format, though it is not entirely clear if the latter should obsolete the earlier H.263 payload format in RFC 2190.
The biggest topic was a continuation of the discussion that began at the previous IETF regarding a generic payload format for the hundreds of existing codecs. Two proposals were presented and several refinements and enhancements were suggested in the discussion that followed. The authors of the two proposals agreed to work together to produce one merged proposal also incorporating the ideas from the discussion.
Two presentations were also given on an RTP payload format and RTCP usage for MPEG4. There are a number of questions still to be resolved regarding conflicts between the RTP and MPEG4 architectures; these are to be discussed further on the mailing list.
Three topics from the mailing list were discussed:
1) a request was made for a means to add application-specific extensions to the RTCP RR packet, but the sentiment was that a new RTCP packet type should be used instead;
2) there were a number of suggestions for redefining the interval over which the RTCP RR "loss fraction" is calculated, but not a consensus for changing it; and
3) it was agreed that the RTP Profile draft should be revised to say no more static payload types would be assigned, but in the meantime a request for an assignment for Qclp should be allowed.
A revision of the generic FEC payload format draft was favorably received given one or two small changes. However, a proposal that RTP should be defined as its own IP protocol received a primarily negative response due to concerns about interoperability problems.
Differentiated Services (diffserv)
The first meeting of Diffserv was attended by more than 250 people. The charter was reviewed with little comment. The strawman Diffserv draft was presented and a series of issues raised on the mailing list were reviewed. Some of these were resolved and some were left for resolution on the list or in the interim meeting in June. A straw poll was strongly in favor of the "code point" approach to DS byte allocation rather than coded bit fields. Another straw poll removed the IN/OUT bit from the specification. An outline of the Diffserv framework document was reviewed, and writing assignments were agreed. Finally four brief presentations on Diffserv-related work were made.
IP Performance Metrics (ippm)
The IPPM session met for one session early afternoon Thursday. Progress was made on two significant aspects of the One-way Delay and the One-way Packet Loss metrics, viz. notes on calibration and agreement on removing the path parameter. There was a report on a second independent implementation of the one-way delay and packet loss metrics from a team at the RIPE NCC. In addition, several interest talks were given.
IP Telephony (iptel)
The iptel group met for the first time as a working group in LA. The agenda was filled mostly with presentations on related technologies to the call processing syntax - CTI call models, sieve mail filter languages, current call processing syntax efforts, and a liaison presentation from ETSI Tiphon. There was also a presentation on the charter, with a focus on the call processing. There has definitely been confusion, even among those who have been following the group's efforts, about what the group is actually working on. (Many people seem to think that the group will be dealing with running telephony over IP networks, like the ITU H.323 protocol, which it does not.) For this reason, some of the presentations were made more informative to ensure that future contributors understood what needs to be done.
Integrated Services (intserv)
The Integrated Services working group did not meet in LA.
Integrated Services over Specific Link Layers (issll)
The ISSLL Working Group met for a single one hour session on Tuesday, March 31st, with attendance of roughly 110 people. Business included: - Announcement of completion of ISATM document set (will be formally sent to IESG as soon as I-D queue reopens after IETF). - Brief status report for IS802 and ISSLOW documents. (Summary: all of these documents not otherwise mentioned below are complete). - Carsten Bormann described revisions to the ISSLOW MCML document requested by PPPEXT, and presented a plan that completes critical-path revisions to the current documents by April 20th. - Andrew Smith presented a revised proposal for IS802 service mappings that addresses a variety of problems raised with the previous version. The following discussion was generally positive but identified some detail to be addressed. A new I-D draft will be submitted immediately following the IETF. - Steve Jackowski presented his planned revision of the ISSLOW service mapping document based on a review by the chairs of the version he submitted before the IETF. A new I-D will be submitted shortly after the IETF.
Multiparty Multimedia Session Control (mmusic)
The MMusic meeting covered three topics: an extended SIP security architecture, SIP call control and the status of SAP security. The SIP security architecture offers (1) encryption for confidentiality and privacy, and (2) authentication for message integrity and access control. SIP call control features have been moved into a separate draft, which will allow the SIP base specification to advance to Last Call while the working group continues to discuss SIP call control. SAP security has been completed and will be merged with the SAP base spec. A review of the WG core protocols (SDP, SIP, SAP, RTSP) revealed that all are in the final stages of completion.
Network Address Translators (nat)
The first ever NAT WG meeting started with a word from the AD on the purpose behind the charter. The WG chairs reviewed the milestones and priorities, then the RFC packaging. There was no dissent. Steve Bellovin gave a presentation and did some Q&A on IPsec, EID's and HostNAT. Tony Hain and Yakov Rekter exchanged views on the IAB NAT draft. It was suggested that Yakov work with the IAB authors on a new version of this draft. It was also suggested that an applicability draft be written. The chairs will continue to work on the basic "what NAT is" draft with the assistance and cooperation of the working group on the email list.
ONC Remote Procedure Call (oncrpc)
Steve Nahm first discussed creating instructions for IANA to follow in assigning RPC program and authentication/security flavor numbers in the future. These instructions are required for IANA to assume this responsibility, and arranging IANA support is a requirement for RPC to continue on the standards track. The consensus was to keep the description simple and possibly recommend the appointment of a RPC consultant/specialist to help with any hard cases the come up. Several topics relating to possible future work in XDR were next discussed. Bill Janssen of Xerox PARC presented ideas relating to international character/string support, more efficient packing of data, reference types and either-endian number representation. Brent Callaghan then discussed suggested improvements in RPC Language data type keywords, addition of a keyword for linked list support, anonymous structures, and for packed data. Both speakers agreed to further develop their ideas and submit proposals to the working group alias.
The group suggested that the current XDR specification (RFC1832) be modified to include a warning in the definition of the "string" data type that certain implementations of XDR do not recognize embedded NULL characters, and some implementations allow ISO-Latin-1 characters in addition to ASCII. Steve Nahm agreed to make these changes and resubmit the draft to the working group.
PSTN and Internet Internetworking (pint)
The first presentation provided the status of the PINT Informational RFC (presented by Hui-Lan Lu, Document Editor); the second one summarized the PINT issues as far as the services and use of SIP and SDP are concerned (presented by Scott Petrack); the third provided the requirements for protocol (presented by Lawrence Conroy); and the fourth provided the outline of the future protocol standards-track RFC. The final WG call for the Informational RFC candidate will be issued a month from now. (No problems are expected.) The major outcome of the meeting was the agreement on the contents of the future protocol RFC (Lawrence Conroy will serve as Document Editor), whose first draft is expected to be published in July. There were no major disagreements.
RSVP Admission Policy (rap)
The group had two presentations on related work: first, a directory schema for policy representation, and second, a stratagem for user-authentication of an RSVP message. The bulk of the meeting focused on the changes and improvements to COPS draft, which is now at version 01. These changes are designed to simplify and clarify the protocol, which was demonstrated by a sample exchange between the client and server. There was also a brief success story concerning the first COPS inter-operability trials. A brief introduction to DIAMETER followed. The session ended with a reminder to send comments on all of the drafts to the mailing list.
Realtime Traffic Flow Measurement (rtfm)
The WG reviewed its two current I-Ds, 'New Attributes' and 'Simple Ruleset Language.' Reports covered preliminary work on TCP attributes, one-way transit time measurements, and the status of the two RTFM implementations. The WG will continue to develop the above I-Ds, and will produce an Applicability Statement for its new 'Architecture' and 'Meter MIB' I-Ds.
Resource Reservation Setup Protocol (rsvp)
The RSVP working group did not meet in LA.
TCP Implementation (tcpimpl)
The TCP Implementation WG met from 9:30--11:30 March 31, 1998 at the 41st IETF meeting. The following items were discussed/decided. The testing tools I-D has failed WG last call. A new section on the ns network simulator and a more detailed security section will be added. The modifications to the known problems I-D were presented. Also, the problems which still need documented were listed and have been categorized into "serious", "not so serious" and "security". A suggestion to split off a separate security issues document has been suggested. It is not clear whether that will happen or not due to underwhelming response when soliciting volunteers.
Amy Hughes presented work regarding re-starting idle TCP connections. A number of mechanisms were outlined. A discussion followed. However, more work is needed before the WG can decide which (if any) mechanism is appropriate to standardize. This should be considered again in Chicago.
Kevin Fall gave a brief presentation on the ns-2 simulator and its new network emulation capability.
Sally Floyd gave a brief overview of the changes to the larger initial window draft and a discussion of the issue followed. Rough consensus was obtained for increasing the initial window size per the current draft (to 2--4 segments, depending on segment size).
TCP Large Windows (tcplw)
The TCP Large Windows working group did not meet in LA.
TCP Over Satellite (tcpsat)
The TCPSAT meeting took place from 9-10:30am at the 41st IETF in LA. The agenda covered status for the group and drafts, final tweaks thought necessary to the standard mechanisms draft, issues without authors for the research issues draft, and some open discussion on the value of creating document describing spoofing being done or contemplated in the global Internet.
The standards mechanism draft should be ready for last call by the end of April but still requires some additional wording giving more detail on how much benefit users can expect from FEC and which environments might not be able to reach "fiber-like" BER quality. After a brief discussion, concerns that Path MTU discovery might not always be good were put to rest after the observation that if the satellite system provider wanted small packets on his link for optimal TCP performance, he would set a small MTU and pMTU discovery would still optimize for the link. There was a brief statement by Jeff Semke on PSC work in auto configuring buffers. The research draft has several new sections and needs authors for several more. If the research issues draft can go to last call before August, TCPSAT will dissolve and will not meet in Chicago.
There was little opposition to the suggestion (originally by Scott Bradner) to support an effort to document TCP spoofing. The goal of this document is twofold: 1) to educate (potential) spoofers about the risks of spoofing and the standards changes (like IPSEC) that will affect their ability to spoof and 2) to educate the IETF what people are (proposing) doing in the Internet. Note that there is wide interest in spoofing in the terrestrial wireless community as well as some in transoceanic fibers. The AD recommended creating a BOF at the August IETF. I-D submissions were solicited and an email list will be created (& a web page). Details will be broadcast to IETF-announce; tcpimpl; tcpsat; and forwarded to end2end-interest mail lists.
Transport Area BOFs:
Explicit Congestion Notification (ecn)
The goal of the ECN BOF was to collect researchers already interested in ECN to discuss outstanding IP, TCP, and UDP-related issues regarding experimenting with ECN, and to introduce the work on ECN to the rest of the IETF community. The BOF was attended by roughly 150-200 people. The agenda of the BOF included an introduction of ECN, a detailed discussion of the additions to IP and to TCP, a report about an implementation of ECN in IPv6, and a discussion of security issues. The meeting ended with a unanimous consensus that the experimental work on ECN should continue.
Multicast-Address Allocation (malloc)
The current version of the charter was discussed. Several topics were raised and judged out of scope by the AD (and apparent consensus): policies for multicast address allocation (setting limits on how many addresses one MASC domain can request) and alternative multicast models (like per-source multicast addresses). Reports were presented on MASC simulation results, AAP implementation plans, and MDHCP implementation experience. Concerns about MDHCP's dependence on DHCP were resolved by agreeing that MDHCP will be compatible with, but not dependent on, DHCP. Therefore, an alternative to MDHCP, MARP, was withdrawn.