2.7 Transport Area

Transport Area Report - Chicago IETF

Working Groups:

Audio/Video Transport (avt)

The AVT working group's primary outputs, the RTP specification and audio/video
profile, have been revised in preparation for advancing to Draft Standard. Changes in
the spec were reviewed at this meeting, with a few changes to be made in one more
round of editing. The main change is to move the discussion of SSRC sampling to a
separate Experimental RFC which will identify IPR claims for one solution. The
profile is less complete; work is needed on the payload type assignment policy and on
registering the encoding names as MIME subtypes.

The WG also has several drafts in process for publication and two drafts about to be
submitted: QCELP payload format and "Guidelines for writers of RTP payload format
specifications". The RTP MIB has been revised a second time and is expected to be in
finished form by the next IETF. Four proposals were presented on multiplexing
multiple RTP streams into one to reduce overhead for specific applications such as
Internet Telephony gateways. The authors of these drafts are to form a common
solution in discussion on the mailing list. A draft revision of the payload format for
redundant audio, RFC 2198, fixes a problem with needing to signal the redundancy
delay in a backward compatible manner. This will be published when going to Draft
Standard. Open issues for the FEC payload format spec were decided at this meeting
so that it can go to last call after splitting out the Reed-Solomon case to a separate draft.
There remain several open issues for the MPEG-4 payload format proposal; the group
agreed to restrict the draft to dealing with Elementary Streams only, and defer
consideration of a FlexMux format until after Version 1 of the MPEG-4 specification
is released and some implementation experience is gained. The meeting closed with
a request to the members to review the proposed new WG charter and milestones.

Differentiated Services (diffserv)

The WG concluded last call comments on the basic header draft and the architecture
draft, which will get one more editorial pass before submission to the IESG. The EF
(expedited) PHB specification was discussed and hopefully will be done before the
December IETF. A preliminary AF (assured) PHB specification was presented and
will be revised with a specific number of queues and drop preferences. Some
simulation results on services using AF were presented and there was a wider
discussion of AF, drop preference, and the associated service characteristics.

A draft on PHB codepoint management was presented but needs more discussion. A
draft on diffserv boundary devices was also introduced and is intended to become an
informational document. Fred Baker agreed to publish a short clarification of router
requirements stating that ICMP replies should embed the incoming DS field unmodified,
to assist in diffserv tracing. Finally there were brief FYI presentations on RTFM and
IPPM measurement methods.
IP Performance Metrics (ippm)

The session began with a status report: framework is now an RFC; connectivity has
passed WG Last Call with only slight changes suggested; 1-way loss and 1-way packet
delay are converging to Last Calls; bulk transport needs arm-twisting; other proposals
are just starting. Most of the meeting was then devoted to calibration error: Matt
Zekauskas presented his additions to the loss and delay drafts to address the issue,
which the audience discussed. There's general consensus on the material, once the
distinction between calibrating the measurement tool vs characterizing jitter in what's
being measured is understood. As soon as this distinction is clarified in the text (within
a few weeks), these drafts will be ready for WG Last Call. Serious concern was raised
about compatibility with fast-moving ITU efforts; our ANSI T1A1.3 liaison (Vern) will
act quickly and the WG Chairs will look for new person(s) to take on the task. The
meeting ended by looking ahead. Most critically, the IPPM metrics need a process that
ensures that they receive as stringent a scrutiny as more traditional kinds of protocols --
as connectivity enters IETF Last Call, we need an understanding with the IESG on the
initial step of this process. Also, draft versions of new metrics must be in the pipeline
by the Orlando meeting or they risk being omitted from the IPPM effort.

IP Telephony (iptel)

The iptel working group met for two hours on Monday, August 24. The main focus
was to resolve the major outstanding issues with the gateway location framework
document. These issues including some basic directions for development of the
protocol, including intra vs. inter-domain operation, single vs. multi-hop, and
aggregation. Consensus was achieved on all of the major items. An additional issue
was raised (by Christian Huitema) which requires serious consideration - that of
assuring that the next hop towards an advertised telephony prefix is actually reachable.
It was also agreed that discussion of some of the other open issues (such as attributes
to describe gateways) should be done on the list.

A revision of the working group charter was presented, including the addition of the
gateway location framework document, and an updating of dates. The dates needed
revision since (1) the dates were set long before the group was approved, and (2) the
order of development had changed.

The group ended with a brief presentation on an update of call processing syntax efforts.
There was some interest, but time ran out before much discussion could take place.

Integrated Services (intserv)

The intserv working group did not meet at the Chicago IETF.

Integrated Services over Specific Link Layers (issll)

The issll group met for one one-hour session in Chicago. The following items were
discussed:

- ATM documents are complete and published as RFC's.

- IS802 documents complete except for service mapping draft. A new revision of
that draft was posted before the Chicago I-D deadline. Raj Yavatkar gave a short
presentation regarding his concerns about the (new) semantics of request merging
specified in this version of the I-D. Discussion revealed that neither the old nor the
new semantics were workable; authors agreed to meet informally during the week
to work on a solution.

- ISSLOW - Issues raised by the PPPext WG regarding the MCML and RTF
documents have been resolved. David Putzoulo presented the status of the iSSLOW
Service Mapping doc. The authors are going to do some significant revision based
on comments from the co-chairs.

- Intserv over Diffserv - question of whether this WG should consider aspects of
supporting Intserv services over Diffserv clouds. Yoram Bernet presented his (&
others) I-D on RSVP and Diffserv. A general discussion followed, with the
conclusion that a large group of those present wanted to see the WG take up this
topic, particularly if the diffserv work is advanced enough by the end of the meeting
that the ISSLL work can build on it quickly.

Multicast-Address Allocation (malloc)

The malloc working group met at IETF 42 on Tuesday, August 25, 1998 (1300-1515).
All protocol specs (MDHCP, AAP, and MASC) have been updated in the last month,
so presentations were made about each of them. Each presentation included a protocol
overview, list of changes since last IETF, and open issues. Other presentations covered
how MZAP fits into malloc work and how advance allocation can be handled. No big
controversies arose and comments were largely confined to known open issues.

Multiparty Multimedia Session Control (mmusic)

Colin Perkins made a brief presentation on changes to the SAP security specification
made as a result of comments on the mailing list. Mark Handley will merge the
security spec with the base spec and submit this to the IESG to become an experimental
standard before the next IETF.

Mark Handley summarized the status of SIP: we have complete WG last call and
submitted the spec to the ADs. The ADs felt that performing IETF last call over the
IETF was a bad idea so delayed that until after the IETF. In the meantime a couple of
errors had shown up in the SIP spec. Henning Schulzrinne described these errors and
the proposed solutions. A new ID will be submitted as soon as the ID editor is
accepting IDs again, and we will ask the ADs to do IETF last call on this revised spec
instead. The consensus in the room was that the changes were small enough that we
didn't need to repeat WG last call.

Colin Perkins presented a proposal for a local message bus for local area conference
coordination, similar to the LBL conference bus, CCCP and PMM. Such a proposal
was first made in MMUSIC at the 1994 Seatle IETF and has been discussed
occasionally since. There seemed to be a lot of interest in this proposal now. It is not
clear that this should be an MMUSIC work item though, as it has the potential to be
used for more general purpose local coordination. The rough consenus was that work
on this should continue, and that we should consider holding a BOF at the Orlando
IETF to see if the interest is general enough to form a new WG.

Network Address Translators (nat)

NAT met Wednesday morning and we went smoothly through the agenda. NAT
terminology had no comment in the meeting and we said it would move to WG last
call. We begged for input for the protocol issues draft. The traditional NAT and DNS
ALG drafts were presented. Tony Hain made a brief presentation of the IAB draft.
Yakov made a heated rebuttal to the IAB draft and there was lively discussion.

Network File System Version 4 (nfsv4)

The morning meeting was structured as a series of presentations. Brent Callaghan
presented brief history of the NFS protocol, and a rationale for NFS version 4 followed
by the working group charter and milestones. Spencer Shepler followed with a
description of the requirements document and a summary of the strawman proposal for
the v4 protocol. Uresh Vahalia of EMC Corp. presented some considerations for large
servers, streaming of data, server advisories, server file copy, and a variety of other
topics. The afternoon session, hosted by Brian Pawlowski, was devoted to an
interactive discussion of protocol issues covering authentication and authorization,
latency and bandwidth, file attributes, and file locking. The session concluded with
an animated debate of issues for DOS and Windows implementations including case
insensitive names, wildcard lookup, file locking semantics, and DOS SHARE
semantics. Brian concluded the session by underlining issues that need to be resolved
by further proposals and discussion on the mailing list.

ONC Remote Procedure Call (oncrpc)

The oncrpc working group did not meet in Chicago.

PSTN and Internet Internetworking (pint)

The PINT WG met on Thurshday, August 27, 1998 from 15:30 to 17:30. The
presentations took place in accordance with the published agenda; one talk on the
PINT Protocol issues, by Henning Schulzrinne, was added, at his request. The issues
presented and discussed concerned: 1) work on the PINT protocol, 2) relation of the
PINT architecture to that accepted by the ITU-T SG 11, 3) revised PINT milestones.
Some issues identified by the presenters (e.g., usage of SIP, the status notification
mechanism, numbering) were resolved at the meeting; the rest were taken for the
on-line discussion. It was noted that the work on the PINT MIB must be completed in
order for the PINT Protocol document to move to Draft Standard. IETF MIB expertise
was solicited.

RSVP Admission Policy (rap)

Base protocol and RSVP extensions to support it are almost ready for last call: some
small clarifications still needed. 4 current documents: {COPS framework, COPS base
protocol, RSVP extensions for policy, COPS usage with RSVP} to be forwarded
together on standards track. Pre-emption Priority policy objects for RSVP still needs
to be written up as a draft (Herzog). Proposed DiffServ extension framework - will
suggest work more on this after RSVP work is done (WG is chartered only for RSVP
right now). Extensions for User Identity policy objects for RSVP may also be worked
then.

Realtime Traffic Flow Measurement (rtfm)

About 120 people attended; 88 signed the list. The WG reviewed its current four I-Ds.
These, together with a new 'Applicability Statement' I-D will be published early
September for WG Last Call. All five will be submitted to IESG in early October.
Two implementation status reports were given, and Nevil Brownlee gave presentations
on NetFlowMet and his SRL compiler.

Resource Reservation Setup Protocol (rsvp)

The RSVP Working Group met twice in Chicago. The first meeting, one hour in
duration, attempted to review the status of all past and in-progress working group
documents. In addition, we heard the status of the OpenGroup standardization of RAPI.
The last topic, reviewing the V1 RSVP spec documents, had to be postponed to the
second session, which was two hours. This second session was largely devoted to
discussion of possible future work on RSVP. There were reports on RSVP extensions
that have been proposed by the MPLS and diff-serv groups, and on the RSVP
aggregation work to date. Lixia also presented her two-level model.

TCP Implementation (tcpimpl)

The tcpimpl WG met on Friday, August 28, 1997 at 9am. We began with a reminder
from Scott Bradner of the requirements for disclosing possible IPR issues when
participating in WG discussions. Vern Paxson reviewed the status of the WG
documents, as follows. The testing tools document is done. The 3 larger initial window
documents have been approved by the IESG and are currently awaiting publication as
RFCs. The re-start document will be updated and is slated to be published as an
experimental RFC. Discussion of the known problems I-D elicited an unusual but
serious bidirectional-transfer implementation problem discovered by Bill Fenner, who
will undertake to write up a description for incorporation into the I-D. Mark Allman
gave a quick review of what we need in terms of a security problems document, which
needs to be worked on energetically or the group will give up on producing it. The
changes made to RFC 2001 were outlined and Sally Floyd gave a quick presentation
on various versions of fixes to TCP Reno. The consensus was that NewReno should be
documented in an experimental RFC. More discussion on the list is needed as to exactly
how to reconcile use of New Reno vs. use of Janey Hoe's somewhat different fast
recovery fix. We discussed the chairs' view that the WG should close after the Orlando
meeting, and did not find contrary opinions. Discussion after the meeting was over,
however, raised the possibility of working on a PMTU discovery document, which
would probably require keeping the group active for at least one more meeting.

TCP Over Satellite (tcpsat)

The TCP over Satellite group met on Wednesday, August 26th, at 3pm. The Internet
Draft "Enhancing TCP Over Satellite Channels using Standard Mechanisms" is nearly
done and should be ready to go to last call on the list for publication as BCP. "Ongoing
TCP Research Related to Satellites" is still incomplete but there are authors assigned
for the unwritten sections. The schedule is to send out a last-call-ready draft before the
next IETF with last call to follow in January 99. It was announced that the plan is for
the next IETF meeting in Orlando to be the last TCPSAT face-to-face meeting. To
continue the TCP spoofing thread from the last meeting, it was noted that the next
steps for those interested would be to produce a short I-D defining terms and
architectures and a second I-D discussing spoofing mechanisms and risks. The web
site is http:tcppep.lerc.nasa.gov/tcppep. The issue was raised as to whether issues
pertaining to TCP over ATM (ABR) over satellite should be included in the research
issues draft; the consensus among the chair, editor, and ADs was that unless specific
TCP mechanisms were at issue, it is out of scope. Dave Beering, Sterling Software,
presented results of a NASA Advanced Communications Technology Satellite
Experiments 118x that achieved 97% utilization of a OC-12 satellite link using a single
TCP connection and Solaris 2.6 w/patch, Solaris 2.7, Windows 4, and Windows 5.
Eric Travis, JPL, presented a summary of the InterPlaNet brainstorming going on at JPL
w/Vint Cerf's.

BOFs:

ss7

The SS7-Internet BOF met for 2 hours August 24th. The co-chairs were Lyndon Ong
of Bay Networks and Nancy Greene of Nortel. There were presentations from a number
of participants on requirements for ss7-internet interworking. The consensus of the
group was that two working groups should be established. These would be one to focus
on transporting IN signaling over IP, and the other to focus on the interface between a
media gateway controller and a media gateway. This is an area where joint work with
other groups, such as the ITU SG11 and ETSI Tiphon, may be valuable.