The RMT Working Group met in Yokohama on July 16th, 2002.
The meeting was opened with a summary report on the WG status presented by Roger
Kermode, followed by update-presentations on
the GRA signaling draft (draft-ietf-rmt-bb-gra-signalling-00.txt),
on WEBRC congestion control (draft-ietf-rmt-bb-webrc-02.txt),
on PGMCC congestion control (draft-ietf-rmt-bb-pgmcc-01.txt)
and on the NORM building block (draft-ietf-rmt-bb-norm-04.txt).
The meeting ended with a discussion on the WG progress and on future plans.
The status-report was given by general protocol areas: most of the ALC-related
drafts are undergoing IESG review following the submission as experimental RFCs.
The only non-ready draft in this area is the congestion control one
(draft-ietf-rmt-bb-webrc-02.txt). Some of the ALC drafts under IESG examination
are also part of the NORM protocol. The main drafts related to NORM are in good
status of advancement (draft-ietf-rmt-bb-norm-04.txt and
draft-ietf-rmt-pi-norm-04.txt). As for the congestion control for NORM, PGMCC
(draft-ietf-rmt-bb-pgmcc-01.txt) is current and almost complete while TFMCC is
expired. However the last submitted version of TFMCC was almost complete. All
the drafts specifically intended for the TRACK protocol are expired and none of
the authors was present at the meeting [more on this later]. Finally the work on
GRA has just restarted and it is still at an early stage.
Tony Speakman discussed the advancement on GRA signaling. The GRA signaling
draft, that will be submitted to the IETF directory after the meeting,
concentrates on a general framework for end-host to request network elements to
execute GRA function. The specific GRA functions are to be defined in separate
specification(s). The presentation was followed by a technical discussion on the
draft content, with two main hilights: the G-bit used to detect the GRA header
in transport packet formats needs to be considered by the other authors of RMT
protocol specifications, and the granularity used by the draft to describe
network-element functions might be revised (made finer). Also the future of the
other two GRA drafts that are expired was discussed. These will not be
resubmitted in the short term, but might be updated at a later stage, with the
intent of providing information on the GRA framework (the intended outcome is
informational RFCs).
Mike Luby presented an update on the WEBRC congestion control. This congestion
control scheme is also discussed in a technical paper to appear in the
proceedings of Sigcomm'2002. NS code will be available by August 1, some
simulations in real networks have already been done but there is a need to do
real-world experiments. The goal is for this to be an RFC by March 2003 at the
earliest.
Gianluca Iannaccone gave a brief update on PGMCC, that was revised to comply
with the 'author guidelines" document and to fix some technical details. Next
step is to develop more on the interaction between PGMCC and feedback
suppression mechanisms. Two additional action-items for the draft authors came
from the discussion: add a description of a possible attack from a malicious
receiver that causes itself to be elected as the "acker" by sending a large
amount of NAKs, but then drives the session rate up by ACKing in an inconsistent
way. The specification should require implementors to protect the network
against this type of attack through heuristics that compare the loss-rate
reported by NAKs and the one inferred from ACKs. The other action item was to
experiment with the protocol in real networks.
Brian Adamson gave updates on the NORM BB draft and presented some simulation
results (the code of his simulation experiments is shared with the reference
implementation that he has). Next steps are: collecting more feedback from the
WG, working on congestion control support (both TFMCC and PGMCC) and GRA
support. There was some feedback on the NORM document structure: Roger Kermode
suggested to revise it to comply to the RMT authors' guidelines, Tony Speakman
to move the content around between the BB and the PI document to improve
modularity and chance of BB reuse.
An evaluation of the WG status an discussion on future action items closed the
meeting.
Most of the work in the charter is close to completion with the exception of the
TRACK protocol and GRA.
As for TRACK, it was not possible to collect information on the authors'
intentions, as none of them was present, however the feeling of the WG was that
the status of the work is probably an indication of lack of interest in pursuing
it further. The plan is to stop the work on TRACK and change the WG charter
accordingly, unless there is a clear indication/commitment from the TRACK
authors.
The work on GRA is being carried on but is substantially behind compared to the
rest, hence the WG will proceed towards the publication of the various BBs and
PIs as Experimental RFC. If GRA is not ready at the time of the publication, the
final integration work will be done in a later revision of the specs.
Finally two possible new working items for the WG were discussed. The first was
a session control protocol, the action item was to form a team to gather
requirements, before making any decision on whether this work should be done in
the WG or not. Mike Luby volunteered to lead the effort with the help of Tony
Speakman and Roger Kermode.
The second proposal was to open the WG to non-one-to-many RMTs. The WG
orientation was not to pursue it at this time. The rationale behind this
decision was to have time to consolidate the results currently achieved and
observe how deployment plays out, before starting to standardize new pieces of
work.
