CURRENT_MEETING_REPORT_
Reported by Peter Ford/Los Alamos National Laboratory
Minutes of the Source Demand Routing Working Group (SDR)
Tony Li opened the meeting and bashed the agenda into shape. It was
subsequently dynamically reordered.
Deborah Estrin gave an overview of SDRP, noting the specification for
SDRP has not changed since the last meeting. She requested that people
read the specification and provide more comments.
Deborah reported on a prototype implementation by Daniel Zapalla at USC
based on the SunOS DARTnet kernel. Tests were conducted on a small
4-node testbed at USC, and on DARTnet. There is a kernel interface
establishing source routes, filtering and encapsulation. There is a
routing socket interface for the D-FIB.
USC is interested in seeing more people pick up the code and build
experimental testbed islands, and then interconnecting them for later
interdomain experimentation.
Christian Huitema asked about anycast, and Yakov noted that the AS
number is a group address. Further work remains to be done in defining
the requirements for anycasting, but there appears to be nothing in the
SDRP specification that would be a substantial limitation.
Tony Li reported on his work on BGP/IDRP interaction with SDRP. The
original idea was to route SDRP control messages to AS representatives
(e.g. to query AS for policy). This functionality is possible by
changing routing of control messages, with the last address being an AS
address. Thus the BGP attribute is no longer needed for this
functionality. Tony will fix the specification to reflect this change.
The IDRP attribute is still useful for tunneling. Yakov has a draft of
this attribute and will post it as an Internet-Draft shortly.
Tony reported on the policy language he has been working on. It is
C-like. Tony would like to get more comments on the specification. The
evaluation of policy is boolean. The functionality will be familiar to
people who are familiar with cisco access lists. The OR operator
(distinct from ||) can be used to ignore a term which cannot be
evaluated (bottom). Christian Huitema suggested that the access control
aspects of this language should be checked against the work on IP
security. Future work will include policy based on source and
destination AS number (using the DNS?), source and destination
communities, and there may be some work on richer pattern matching on
the entire SDRP route if there is a need. There could be some work on
time varying characteristics, such as load or delay. Steve Hotz sent
mail suggesting that Tony think about evaluating policy terms to
preference continuous values instead of boolean values.
Tony reported on his work on Information Distribution. The plan is to
get the current AS topology from some static source. It was suggested
to see RIPE-81 for an expression of this information. Christian noted
that we probably need to consider higher levels of aggregation of this
information above ASs.
Deborah stated that she needs to do a rework of the futures document
which deals with scaling issues. She will update the document.
Deborah reported on SDRP setup work. Setup is done via an explicitly
source routed packet with the probe bit set. The motivation for these
setups is to reduce the header size. It is not a requirement for a
router to participate. It can strip the probe bit, send it ahead, and
send a setup rejected message back to the originator.
There was a question if the routers have to maintain information on
source routes that it is currently setup for: the answer was yes. It
was again noted that this is soft state.
There is a draft specification of SDRP Setup that will be distributed
later in the summer.
Christian stated that setup needs to be investigated to see what its
interactions are with regard to load splitting.
Deborah noted additional work that will be considered, including
multicast (and its relationship to ESL). This would be used to establish
branches on multicast trees. There also needs to be better tools for
building SDRP routes.
Task List:
1) Review of Specification. Tony will lead a section by section read
of the document at the next IETF. Christian Huitema, Keith Mitchell,
Shezahd Merchant agreed to read the entire specification and comment
as soon as possible.
2) MIB -- Yakov will find someone to do this.
3) Usage Internet-Draft -- Peter Ford
4) Gated hack needed to populate D-FIB. Sue Hares/John Scudder
5) Experiments with SDRP
Merit -- Hares and Scudder
Lothberg
Scott Brim - maybe
6) Mapping packets/ packet classification
7) DNS support for IP to AS number mapping
Peter Ford will talk with RIPE-81 people
8) Policy Language -- Tony Li and Steve hotz
9) Request/Response protocol -- Tony Li
10) Futures Update -- Deborah
11) Tony will do two specification changes for:
passing control messages to ASs
setup -- translate request to explicit route
Attendees
Michael Anello
[email protected]
Toshiya Asaba
[email protected]
Cynthia Bagwell
[email protected]
Dennis Baker
[email protected]
John Ballard
[email protected]
John Burnett
[email protected]
Ross Callon
[email protected]
Henry Clark
[email protected]
Francis Dupont
[email protected]
Deborah Estrin
[email protected]
Dino Farinacci
[email protected]
Peter Ford
[email protected]
Vince Fuller
[email protected]
Susan Hares
[email protected]
Denise Heagerty
[email protected]
Frank Hoffmann
[email protected]
Nandor Horvath
[email protected]
David Jacobson
[email protected]
John Krawczyk
[email protected]
Tony Li
[email protected]
Robin Littlefield
[email protected]
Jun Matsukata
[email protected]
Keith Mitchell
[email protected]
Ramin Najmabadi
[email protected]
Peder Chr. Noergaard
[email protected]
Erik Nordmark
[email protected]
Petri Ojala
[email protected]
Juergen Rauschenbach
[email protected]
James Reeves
[email protected]
Yakov Rekhter
[email protected]
Duncan Rogerson
[email protected]
Shawn Routhier
[email protected]
John Scudder
[email protected]
John Stewart
[email protected]
Fumio Teraoka
[email protected]
Paul Traina
[email protected]
Rene van der Hauw
[email protected]
Willem van der Scheun
[email protected]
Jost Weinmiller
[email protected]
Jessica Yu
[email protected]