home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Internet Info 1997 December
/
Internet_Info_CD-ROM_Walnut_Creek_December_1997.iso
/
ietf
/
cip
/
cip-minutes-90may.txt
< prev
next >
Wrap
Text File
|
1993-02-17
|
5KB
|
108 lines
CURRENT_MEETING_REPORT_
Reported by Claudio Topolcic/ BBN
The CO-IP Working Group met at the May 1-4 IETF Meeting at
Carnegie-Mellon University. During the Tuesday sessions we tried to
pick up where we had left off in Florida State. We heard updates on
DARTNet and the TWBNet. Tony Mazraani gave a progress report on the
COIP kernel and a presentation on Washington University's work on
Resource Management in Broadcast Lans. Work toward defining experiments
for the DARTNet was hindered since not all the key people were present.
We spent the balance of the sessions discussing the current draft of the
ST-2 protocol specification.
Charlie Lynn had previously edited and distributed the current draft of
the ST-2 protocol specification. He had also written up a number of
issues that needed more thinking. The group discussed these issues and
a few others that came up during the meetings.
A number of editorial comments to the draft were discussed. These
included some minor restructuring to minimize repetition and increase
clarity. More forward and backward pointers were suggested, as well as
more examples. Numerous editing changes were suggested.
We discussed the relation between ST and IP. We decided to allow two
forms of the ST header. The short form is as had previously been
specified. A long form is structured like an IP header so that it can
be processed by IP-only agents, and takes the place of the concept of IP
encapsulation. The long form may also be used when IP security is
required or to reduce either deliberate or accidental denial of service
problems.
The issue of use of multicast lead to a lot of discussion. Ideally, we
would like to be able for an ST agent to request that the local network
dynamically create a multicast group for use by a stream, as its use
could reduce the network bandwidth required to support the stream.
Unfortunately, there does not seem to be much support for dynamic
management of multicast addresses (how does a ``user'' dynamically
request a multicast address at a given protocol layer, what agent(s) on
a network robustly assign multicast addresses, how are the assigned
addresses mapped into addresses for use above the network layer, e.g.,
IP multicast addresses, how are the assigned addresses reliably
released/garbage collected, etc.).
It was felt that trying to create such a service was a challenging
problem tangential to the work of the Working Group and should be
delegated to some other group. The result was to either to use
replication instead of multicast, or to use static multicast groups.
The problem with the former is wasted bandwidth, that with the latter is
scaling -- what were formerly seperable problems (solvable by each
stream independantly) now become problems which must be solved in common
by all streams on a network. HID negotiation is one example.
1
We discussed mechanisms by which changes could be made to established
streams. For example, it may be desirable to allow a request to change
a stream's bandwidth to allow a range of possible bandwidths. Also,
when a new target is added to a stream, it would be desirable to
decrease the stream's bandwidth, if that is allowed by the stream, if
the new target can only be added with that decreased bandwidth. These
features causes some difficulties in coordinating the changes among the
ST agents, as well as the applications, while maintaining the
uninterrupted flow of data packets.
Other specific issues discussed included the following:
1. A Target cannot be an IP multicast group.
2. The ACCEPT message should be delayed until the HID negotiation has
been completed.
3. We are not addressing the issues of spoofing (beyond the security
features to be provided for IP by SDNS), intentional denial of
service, or unintentional denial of service resulting from broken
routes.
4. The structure of the ``Group of Streams'' specification.
5. Whether source routes would be strict, loose, strict in ST and
loose in IP, or something else. This issue was not resolved.
ATTENDEES
Fred Bohle fab@saturn.acc.com
Terry Braun tob@kinetics.com
Stephen Casner casner@isi.edu
Danny Cohen cohen@isi.edu
Richard Fox sytek!rfox@sun.com
Jonathan Goldick goldick@b.psc.edv
Jack Hahn hahn@umd5.umd.edu
Charles Lynn clynn@bbn.com
Tony Mazraani tonym@flora.wustl.edu
Zaw-Sing Su zsu@sri.com
Ian Thomas ian@chipcom.com
Claudio Topolcic topolcic@bbn.com
Dave Wood wood@gateway.mitre.org
2