2.7 Transport Area

Transport Area Report - 46th IETF - Washington DC - November 1999

Working groups:

Audio/Video Transport (avt):

This meeting marked the completion of working group last call on the set of drafts that have been prepared for advancement of RTP specification and A/V profile to Draft Standard status. The working group accepted the changes that had been made in the drafts posted for last call as well as the few changes proposed in response to comments received during the last call. With those changes, the working group agreed that the drafts should be submitted for IESG Last Call. This will be done immediately.

Several RTP payload format specifications are also progressing. The Telephony Tones draft completed working group last call before this meeting and will be submitted for publication after a final revision is posted. The five payload format specs for Pointers, DV Video, DV Audio, Loss-tolerant MP3 audio and Conversational Text were all accepted for working group last call at this meeting. Payload formats for AAC audio, AC3 audio and HDTV video, plus a proposed clock rate modification for MPEG 2 Transport Stream payload format, were presented but require more development and/or testing before submission.

Several MPEG committee members attended this meeting to discuss the three payload format proposals for MPEG-4 media. A design team meeting was held in addition to the two AVT sessions. The WG concluded that multiple payload formats will be required for different application scenarios. Two proposals will be submitted for publication as experimental protocols, and one new proposal will be submitted for standards track after further review and comment by the working group.

The discussion of multiplexing continued at this meeting. It appears that the GeRM proposal has not garned the necessary support. A new payload format for Nx64 multiplexed audio was presented, but there were multiple comments that this should be restricted to a pure circuit emulation payload format. The need for multiplexing may instead be satisfied by a combination of RTP header compression with multiplexing of PPP packets and tunneling the result over IP. Three enhancements to RTP header compression were presented to facilitate this mode of operation.

With the submission of the RTP spec package, it is time to revise the working group charter. Most of the current work should be completed in another one or two meetings. However, a new activity was proposed at this meeting to develop RTCP-based retransmission for unicast RTP. The Area Director says this work is within scope, so the details of what work is needed in AVT and what coordination with other working groups (for example, on congestion management) will be worked out.

Differentiated Services (diffserv):

Diffserv met twice in Washington with the usual attendance of around 400 people. The first session was devoted to mainstream WG items:

The second meeting briefly reviewed eleven unsolicited drafts, looking for expressions of WG interest. In such a busy session it was only possible to make a superficial judgement, but authors were encouraged to start discussion on the list for drafts where there seemed to be some interest.

Endpoint Congestion Management (ecm):

The ECM working group met for the first time on Tuesday, Nov. 9. The meeting began with an overview of the charter, which states two main goals: (1) to provide a standard set of congestion control algorithms that transport protocols can take advantage of rather than having to develop their own, and (2) to develop mechanisms for unifying congestion control across an appropriate subset of an endpoint's active unicast connections. A key point is that determining the granularity of ``common destination'' and deciding when to bundle connections together are hard problems for which the WG will look to the IRTF for active research. Some skepticism was expressed as to whether the milestones are realistic, but the chair defended them by arguing that drafts exist for two or perhaps three of them, and the final one does not appear that difficult.

The meeting's main goal was to get all of the WG members on the same page, and it appeared to succeed in that regard. There was lengthy discussion of numerous aspects of the group's core document, draft-balakrishnan-cm-01.txt, and a revision reflecting the comments is expected soon. One particular concern was that this document will be defining much of how congestion control evolves in the future, and thus requires especially close scrutiny; however, a counter-argument to this viewpoint was made that because the architecture requires deployment only at the sender, the installed base is much easier to upgrade than in the case when both sender and receiver must be upgraded.

The other document discussed was draft-floyd-cong-00.txt. It elicited some discussion over whether particular principles discussed in the document are in fact valid, given realities concerning restricted bandwidth to the home. The counter-argument that, even if this restriction dominates congestion patterns today, one still shouldn't assume that this is some sort of invariant pattern, appeared to satisfactorily address this concern, and the draft document appears to enjoy good support.

IP Performance Metrics (ippm):

The ippm working group did not meet in DC.

IP Telephony (iptel):

The iptel working group met for a single session. The WG reviewed the TRIP framework document, and has also recently received comments from an IESG review. These will be folded in (all minor) and the document resubmitted for informational (it was previously submitted and withdrawn for a protocol name change). The CPL framework was discussed, and a working group last call issued on it. TRIP was then discussed. The main open issue is I-TRIP (the equivalent of I-BGP for TRIP). Several synchronization algorithms were proposed. Consensus seemed to be to stick with what is safe, and borrow the mechanism from IS-IS. As the controversial attributes, capacity and pricing, have been moved to separate documents, many of the remaining open issues disappeared. The chairs anticipate another one or two revs and then a working group last call after the next IETF meeting.

Integrated Services (intserv):

The intserv Working Group did not meet in DC.

Integrated Services over Specific Link Layers (issll):

The ISSLL WG met for one session on Thursday, November 11th from 1530-1730. The status of the current set of documents was presented; nearly all of them are ready to go to the IESG. Next, a quick discussion took place about DSCP mapping notifications. It was agreed that this is a problem but shouldn't hold up the current documents. The group spent the rest of the meeting on service mappings for DiffServ over IntServ. This was very constructive with lots of input and discussion. Consensus on the general approach of mapping Controlled Load to AF and mapping Guaranteed to EF was reached. A draft will be forthcoming.

Media Gateway Control (megaco):

The megaco working group met twice, once on Wednesday and also on Friday.

The Wednesday meeting opened with the announcement that the IETF is still tinkering with process. Given that the Megaco protocol still has some deficiencies, the latest proposal is to mark the completion of the work by going to IESG last call and issuing a Proposed Standard BEFORE the ITU-T decision meeting in February. The intent is to achieve a cleaner handoff to Study Group 16. ITU-T management had been advised of this change in procedure and had agreed with it. Once H.248 has been decided, further development can revert to the original plan.

At the Wednesday meeting, Tom Taylor presented a review of the protocol against the megaco requirements document. A number of discrepancies were noted, some of which raise issues for the protocol. In a few cases the meeting agreed that the protocol had met intent (@@@ "met intent"?) and the requirements needed clarification. Many cases were noted where the requirements will be met when work on package specification has been completed. (As a result of a decision made at the IETF plenary on Wednesday night, the requirement to support legal interception will be deleted from the document.) Brian Rosen presented the results of the ITU SG 16 Redbank meeting.

A preliminary draft for a Megaco MIB was discussed at the Friday meeting. Most of the remainder of the meeting was spent discussing open issues, and particularly proposals arising out of the design meetings and informal discussion the night before. The meeting concluded with a discussion of the future of Megaco with no specific conclusions being reached.

Multicast-Address Allocation (malloc):

The Multicast Address Allocation Working Group (MALLOC) met 0900-1130, Friday, November 12, 1999. Brief presentations covered status for the Architecture, AAP, MADCAP, MASC, and Scope Nesting documents. MADCAP has been approved for Proposed Standard. The other documents should be going to the IESG soon. Several useful, though relatively minor, comments were made. Dave Thaler gave a talk about zero-configuration multicast address allocation (draft-thaler-zeroconf-multicast-00.txt). There was lots of interesting discussion (generally positive).

Multiparty Multimedia Session Control (mmusic):

The MMUSIC WG met once during the 46th IETF, for a two hour slot. The meeting was chaired by Joerg Ott. Some 150 participants attended the meeting. After a brief review of the agenda, Joerg gave an overview of the tentative new charter of the Working Group. He noted that Ruth Lang, Eve Schooler, and Mark Handley have stepped down as co-chairs and then introduced the tentative work items. The new ones were the development of a capability description framework as a focus, and the design of a local coordination mechanism for conferencing application (the Message Bus). Further work items include collecting and documenting existing and emerging extensions to SDP (Session Description Protocol), moving SAP (Session Announcement Protocol) ahead to Proposed Standard, revising the Real-Time Streaming Protocol (RTSP) and developing a simple conference control protocol for tightly coupled sessions. Finally, he announced that a WG Last Call will be issued on the IETF Multimedia Conferencing Architecture I-D shortly after this IETF.

Colin Perkins briefly discussed recent work on version 2 of the Session Announcement Protocol. He pointed out that the current SAP should be documented as Experimental RFC (rather than standards track) because of the well-known scaling problems. A successor version of SAP should then be targeted at Proposed Standard. A number of comments received during the presentation will be incorporated into the draft; the I-D will be re-submitted shortly after the IETF and then a WG Last Call issued. Colin Perkins also introduced extensions to SDP to deal with administrative scoping. It was debated whether this may be an issue of SAP rather than SIP. No conclusion was reached on this subject and investigations will continue.

Glenn Parsons presented extensions to the SDP attributes to support T.38 Real-time fax over IP with SIP as signaling protocol. He noted that ITU-T SG8 plans to incorporate these extensions as a new Annex to T.38, and that this document will be decided in the February 2000 meeting of ITU-T SG8. Comments on the work should be sent to the mailing list and will be summarized as input to the ITU-T meeting by January, 19th, 2000.

Andrea Basso presented an approach towards MPEG-4 stream control with RTSP, particularly the issue of initially retrieving a scene description from an RTSP server. It was pointed out that similar work has been presented in MMUSIC before, and the respective presenters agreed to join their efforts to develop a solution. A corresponding I-D will be submitted within the next two months.

Joerg Ott provided an update of the work on the Message Bus: the transport is deemed stable by now, message semantics for call and media engine control are progressing. The Message Bus transport is targeted for Proposed Standard, first revisions of semantics documents may only aim at Experimental RFC. It was noted that the some of the messages presented may be viewed as some kind of an API and hence may not be subject to standardization in the IETF. The chair agreed to consult the Area Directors on this matter. Revised drafts of the Message Bus documents are to appear shortly after the IETF.

Finally, Joerg Ott presented the current status of the work on the capability framework, particularly outlining a system model and reviewing requirements. The basic concepts were agreed upon. Joerg also raised the issue of learning and re-using as much as possible from the work of the CONNEG WG; comments were made that requirements were largely similar and people urged not to develop new solutions except where absolutely necessary. This approach was agreed upon and the authors will continue with further detailed investigations of the CONNEG work.

Network Address Translators (nat):

The NAT Applications Guide will get one more draft and then probably go to last call. The SNMP ALG received comments regarding support of SNMP versions 1, 2 and 3 and it will get another rev. The NAT Interface Framework draft did not receive any comments. There were many comments on the main RSIP drafts. One particular issue discussed was the use of TCP versus UDP for RSIP. The rough consensus of the room was to support TCP only for now. The 3 non-WG RSIP drafts received critical review and will likely be modified significantly in the future.

Network File System Version 4 (nfsv4):

Brent Callaghan welcomed attendees to the 5th meeting of the NFSv4 working group and introduced document editor, Spencer Shepler, who gave an overview of recent changes to the protocol spec document. The additions to the 02 version of the spec were minor, consisting of updates to some of the file attributes, changes to create for symlink and special files, some changes to the OPEN operation, and updates to READDIR.

Brent gave a short report on the NFSv4 "bakeoff" that was held in San Jose Oct 19-21. At this event, six implementations of the 01 spec were tested for interoperabity.

Mike Eisler followed with several topics. LINK and RENAME have been changed to take a second filehandle from the saved filehandle. He then provided some detail on a minor versioning proposal that provides a framework for clients and servers to assert their minor version support as well as rules for the types of changes that should be allowed in a minor version. There were some questions as to the value of minor versioning (c.f. the RPC version number) though consensus seemed to indicate that minor versioning was worthwhile. There was considerable discussion of the issue of compatibility between minor versions, negotiating minor version levels, and whether implementing different versions should be mandatory or optional.

Mike also described the current state of LIPKEY authentication and asserted its importance as a security mechanism for Internet use of the

Dave Noveck's presentation covered some of the problems that could arise from the stateful nature of the NFSv4 protocol due to the inclusion of file open state, lock state, and delegation state. The presence of leases implies that clients and servers have to pay attention to lease timeouts. Space management is harder and the server could have problems due to unbounded use of resources.

Brian Pawlowski concluded the meeting with a presentation followed by some open discussion. He presented a slide with results that supported the use of TCP (over UDP) as a primary transport for NFSv4. He listed some of the items yet to be added to the spec and suggested that the spec be declared functionally complete when these items are added. The Q&A session heard questions on implementation issues with concurrent NFS and local use of files and volatile filehandles.

ONC Remote Procedure Call (oncrpc):

The oncrpc working group did not meet in DC.

PSTN and Internet Internetworking (pint):

The pint working group did not meet in DC.

Performance Implications of Link Characteristics (pilc):

The PILC working group met at IETF 46, and reviewed four of the five workgroup deliverable I-Ds.

SLOW (slow links) received relatively few comments.

ERROR (high error rates) will have an applicability statement added in the next revision, to advise readers as to whether they need to apply these mitigations.

PEP (performance-enhancing proxies) version 01 was submitted late, so the authors presented the changes that will be in the next announced version. This document needs editorial attention.

APATH (Asymmetric paths) generated the most comments - in particular, the comment was made that TCP handles a high level of asymmetry today, and it's not clear that real networks have a higher degree of asymmetry than TCP can handle. This will be investigated further by the chairs - the document authors have done a thorough job of surveying the literature, but it's important to make sure that a real problem is being addressed and that a spec indeed needs to be written.

LINK (guidance to link-layer designers) was submitted in time for DC, but the principle author was unable to attend this IETF meeting. Because this draft was just announced in the flurry of pre-meeting announcements, the chairs deemed it acceptable to have discussion of the draft begin on the list.

The chairs expect to put SLOW and ERROR, and possibly APATH, forward for WG last call in the near future. PEP and LINK will probably take longer.

Realtime Traffic Flow Measurement (rtfm):

The rtfm working group did not meet in DC.

Reliable Multicast Transport (rmt):

The Reliable Multicast Transport Working Group met for two hours on Thursday, November 11. The meeting started with presentation of the working outline of the Guidelines for Authors of Building Block (BB) and Protocol Instantiation (PI) Specifications. As anticipated, this guideline draft is very much needed: several of the ensuing presentations blurred the lines between BB and PI. The draft was promised for the weeks after the IETF meeting, to be written by the working group chairs. The agenda was packed because due to a scheduling glitch the WG did not receive the second requested meeting slot. The chairs announced that the series of brief presentations for the rest of the meeting should be viewed partly as solicitations for interested volunteers, who should identify themselves to the working group chairs.

Mark Handley gave a brief update on the congestion control BB, reporting that the RMRG will be sending a specification for sender-based congestion control to the RMT WG in about six months. There needs to be an assessment of when a receiver-based congestion control BB will be ready. In the next presentation, Tony Speakman introduced the Generic Router Assist BB. GRA is a filter approach, with non-combinable, small numbers of filter types. Work is needed to flesh out application to varied protocol families and mechanisms.

Three presentations followed to help organize the initial efforts on the three PI families. The relevant working group drafts have not been started yet. NACK-based was presented by Brian Adamson, Tree-Based ACK by Brian Whetten, and Open Loop by Michael Luby. One thread of discussion about NACK and TRACK was the sense that the BB/PI boundaries in them needed to be clearer. Whetten asked participants to try to meet informally before the end of IETF to have more discussion about TRACK. Luby proposed renaming the Open Loop family Asynchronous Layered Coding (ALC). There was little time for discussion of this suggestion, or of a scheme that he has developed involving mechanisms in this PI family combined with source-only PIM. See the slides and full minutes for details of these three and the other presentations.

The last technical presentation was on the security building block or blocks, presented by Thomas Hardjono, speaking as both the co-chair of the SMuG (Secure Multicast) Research Group and a working group contributor. There was not yet consensus on RMT's security deliverables.

There was consensus that it would be worth having an interim meeting about all of our topics. The final item was a brief preview of the revised charter for RMT, which is in negotiation with the Area Directors.

Resource Allocation Protocol (rap):

Andrew Smith, the rap working group chair, was unavailable for the meeting, so he asked Mark Stevens to chair the meeting. Scott Bradner gave the status of the batch of rap documents that had been last-called a while back. There were a few wording issues that arose out of IESG review. Changes to the documents were made to fix the issues and the documents will be ready for publication as soon as the Internet-Draft process is re-enabled after the IETF meeting.

Yoram Bernet presented his draft to add application and sub-application identity policy element for use with RSVP. The meeting participants offered no comments in opposition. David Durham gave a status update on the COPS specification draft. MTI Security requirements added to the draft. Working group and IESG last calls are complete. Next, the draft is to be submitted to the RFC editor.

David Durham also presented draft on COPS usage for policy provisioning. The presentation was received with little debate.

Yasusi Kanada presented an alternative device MIB for quality of service. The central message seemed to be that his alternative MIB provided downloaded applications with better control over devices. Meeting participants offered questions that indicated that there was some confusion as to how the alternative MIB made it easier to offer quality of service in networks. Some participants questioned the viability of downloaded applications in the environment described. In the interest of time, the acting chair asked the presenter and participants to continue discussion on the mailing list. At the close of the meeting the chair updated the group in the status of various working group documents. Of particular interest is plan to go to working group last call on the COPS usage I-D and the COPS client MIB immediately after the meeting.

The working group also plans to submit the COPS Usage for RSVP document to IESG for publication as an RFC in January 2000, go to last call on the revised version of the Object Syntax I-D after incorporating mailing list discussions, and submit the Object Syntax I-D to the IESG for consideration as RFC in March 2000.

Resource Reservation Setup Protocol (rsvp):

The RSVP WG met for the VERY last time in Washington, DC. The meeting discussed two issues.

Silvano Gai of Cisco gave a short informational presentation on his draft on RSVP proxies. There was no opposition to move this work forward through the RAP WG, with a joint last call with RSVP. It uses RSVP in ways that were not anticipated by the int-serv work, but it defines no new protocol, so it is expected to become an Informational RFC.

The second topic, which occupied the majority of the session, concerned the proposed changes for performance and liveness. Lou Berger reported on recent changes to -refresh-reduct-01.txt. Franco Tommasi summarized the difference between their proposal and the Berger ID. George Swallow presented a hierarchical approach to further refresh reduction. After a difficult and long discussion of technical details, the WG gave general agreement to moving this work forward based mostly on the Berger draft. It was decided to recommend local checksumming of state to (partially) prevent getting out of synch, and there was a desire to keep it as simple as possible while accomplishing the desired goals. The intent is to revise the Berger draft as soon as possible and move it to Proposed Standard. Bob Braden agreed to be added as another author and to do a bit more work on the document.

Service in the PSTN/IN Requesting InTernet Service (spirits):

The SPIRITS working group meeting went well, though somewhat unfocused. Volunteer editors were identified for the two informational RFCs; the current practices RFC will likely include a section on relevant existing or in-progress APIs. (API development was brought up by some speakers, but was ruled out-of-scope for SPIRITS.) The protocol RFC may be split into two pieces, just because of size.

There was the usual concern expressed over access to ITU documents. The chair noted that that's an ongoing concern, and probably best not fought out in the SPIRITS working group. There is also pressure on the boundaries, especially with VoIP and PINT; but these pressures seem manageable.

Session Initiation Protocol (sip):

SIP met for a whole day interim meeting and two slots during IETF itself. The interim meeting focused on the DCS proposal, out of Packetcable. DCS is based on SIP, but proposes a number of modifications to SIP (and RSVP) to support carrier class telephony. Packetcable is interested in having their extensions standardized within IETF. Half the time was spent reviewing the DCS proposals, and the other half on discussions. The basic conclusion was that people, for the most part, accepted the requirements that were proposed. However, their solutions are based on telephony only in a single provider network, assuming a user-trusts-the-network security model. SIP as a general protocol has a broader set of operating ranges. As there is much useful things in the DCS proposals, the general consensus appeared to
(1) review their requirements and decide which ones to accept; it is likely most or all will be accepted,
(2) take their extensions as a starting point,
(3) break them into component extensions/modifications, and
(4) see what, if anything, needs to be changed to be compatible with SIP and work in the more broader SIP context, while solving their needs.

The Monday meeting discussed legacy signaling and SIP. Several presentations related to what is now known as SIP-T (aka SIP-BCP-T), formerly known as SIP+. SIP+ is nothing more than carrying ISUP payloads in SIP. The consensus of the group that this work was interesting, but would need to be worked in such a way that it was functional multi-hop and multi-provider without imposing restrictions on general SIP usage. Otherwise, SIP is just being used as an ISUP transfer protocol, which it is not. There appeared to be consensus to work on SIP-T, but it wasn't clear whether this would be done in sip, in a new group, or whether it would proceed as proposed, experimental, or informational. This will be discussed further with the ADs. A proposal for Q.SIG (a legacy inter-PBX signaling protocol, mainly used for inter-PBX features) over SIP was rejected, since it invokes features solely based on Q.SIG, not on the SIP semantics. In this case, SIP is not doing anything more than point to point transfer of data. A presentation was made on call-flow examples with SIP. This is a fairly lengthy educational document that shows call flows with full message examples. It was agreed this work was very valuable, and would be extended to include additional examples (such as the "torture tests" from the bakeoff) and proceed as a work item, as an informational RFC.

The Wednesday meeting started with work currently in the charter. Work on session timer, 183 provisional, and caller preferences is progressing well. The only open issue is what the status of the 183 work will be; in particular, how much is folded into the next rev of SIP, and how much is standalone. A draft was presented that discusses how SIP, RSVP, OSP and COPS are used together for a complete solution for IP telephony. There was much interest in this kind of architectural work, but it wasn't clear where it should be done. There were, among others, proposals for an informational RFC outside of any group, and possibly even a new working group. AD guidance is needed on this issue. A presentation was made on transport of DTMF within SIP. This is a controversial issue and no consensus was reached. Drafts on using SIP for MPLS setup, and for mobility, were presented. There did not appear consensus to work on these. Finally, a draft was presented on SIP servlets, an API for SIP on proxies. Several people were interested in this (the chairs are aware of numerous independent efforts on SIP servlets). Scott Bradner's input was that IETF is not the best place for this work. There was agreement to set up a mailing list to discuss this to determine where, IETF or otherwise, this work should proceed.

Signaling Transport (sigtran):

The SIGTRAN Working Group met on November 9th, with ~180 persons attending. The group reviewed the current status of work on the SIGTRAN charter. Of the original milestones, the requirements/architecture document has been completed and issued as RFC 2719. Four protocol documents are in draft form and need to progressed as rapidly as possible. One document, SCTP specification, was agreed to be progressed to Working Group Last Call based on the current draft, <draft-ietf-sigtran-sctp-03.txt>. Three others, adaptation protocols, will be targeted for Last Call in December, assuming they are ready. The group reviewed related work in PacketCable, ITU-T SG16, ITU-T SG11 and 3GPP and found interest in these bodies to utilize some of the SIGTRAN results. Finally, some proposals for new work were discussed. Work has being initiated on protocol MIBs. There was also support for work on Applicability Statements, but not enough progress was made on identifying specifically what would be addressed, so more discussion was felt to be necessary on the list. If there is sufficient interest, a request will be made for an extension to the group's charter.

TCP Implementation (tcpimpl):

The tcpimpl working group did not meet in DC.

TCP Over Satellite (tcpsat):

The tcpsat working group did not meet in DC.

Telephone Number Mapping (enum):

At its first meeting, the ENUM working group reviewed the basic charter and scope, which is to map from telephone numbers into a set of URLs (or the information contained in set of URLs). ENUM will concentrate on developing a DNS-based solution for mapping queries and responses: the "back-end" services, such as administrative questions of how the data gets into the DNS, is outside the scope of ENUM.

Patrik Faltstrom presented his I-D draft-faltstrom-e164-03.txt, and the group discussed the suitability of this draft as a basis for ENUM's work. It appeared that there was rough consensus that the approach taken in the draft is reasonable enough, and a new draft, incorporating comments from the WG meeting, will be prepared. The draft should at the very least be reworked to present a more generic mapping service, rather than a set of examples narrowly focusssed on implementing number portability.

Transport Area Working Group (tsvwg):

The TSVWG working group met for the first time on Monday, Nov. 8. Scott Bradner introduced the working group, explaining that its role is to serve as a home for transport-related efforts that are too small to merit their own working group, yet don't fit within an existing working group's charter, either.

Five documents were discussed. The first concerns emending RFC 793 to state that a TCP should ignore the IP Precedence field in packets it receives, rather than terminating connections for which there is a precedence mismatch. This change accommodates both existing practice of altering the precedence field in flight, and the Diffserv architecture, which assumes that those bits can be altered in flight. The draft was well received, particularly when it was pointed out that there is a well-established practice of ignoring that part of RFC 793.

A document to enhance TCP SACK reporting to aid in the detection of redundant transmissions, to be published as Experimental, was likewise well received, and another Experimental, TCP "rate halving", did not garner any objections. A document on TCP "congestion window validation" also received rough consensus for publication as Experimental, though with some questioning as to the setting of a particular constant in the draft. Discussion of this point then turned into a general discussion of the role of Experimental documents; it was noted that the point isn't to propose a scientific experiment with particular variables and controls, but rather an engineering experiment.

The final document discussed is a codification of the existing de facto standard for computing TCP's retransmit timeout. The main discussion point concerned the minimum value on the timeout of 1 second, which some view as overly conservative. It appeared clear that good consensus exists for keeping it. The other change noted for the document is to specify that more conservative algorithms than that discussed in the document should also be considered standard-conformant.


Deployment Considerations of Implementing Differentiated Services (decides) This was a follow-up to the BOF held in Oslo at the 4th IETF. At the first BOF there was consensus to hold another BOF, but not one for a working group (yet). The purpose of this BOF was to allow work since then to be presented, and to then see if there were grounds for pursuing a working group, or else providing input to existing (or other proposed) working groups.

The outcome of the meeting was that another BOF is not needed, nor is a working group. However, it was felt that:

The community needs to continue to refine the traffic source models and information about traffic matrices (e.g. for the TE wg) so that simulations and analytical work are kept realistic and relevant to the IETF.

Also needed is a path for measurement results to be reported to the various relevant IETF working groups. IPPM and Diff-serv and others need to liase w.r.t this.

It is possible that soon, but not quite yet, an industrial forum for deploying class and quality of services may be needed. The IESG might consider a directorate perhaps sometime in the next year.

There is a proposed IRTF research group that can help with the first of these tasks to some extent. The expectation is that new WGs will deal with more interim matters, such as composing edge-to-edge services from PHBs.