Internet Standards/CD1.mdf/INET/IETF/ISIS/94JUL.MIN

Text File | 1994-08-08 | 3.6 KB | 78 lines

Editor's note:  These minutes have not been edited.

Date: Thu, 28 Jul 1994 18:58:18 -0500
From: Ross Callon <[e-mail]>
Subject: IS-IS working group minutes


IS-IS Working Group, Wednesday 1:30
Ross Callon is resigning as working group co-chair. Tentatively (subject to
management approval) Doug Montgomery will replace him. Chris Gunner will
continue as co-chair.

Doug presented the documents related to pushing I.IS-IS to Draft Standard.
It is suggested that the MIB go to proposed standard, and I.IS-IS
("son-of-RFC1195") go to Draft Standard. The protocol experience and
protocol analysis needs to be issued (as informational RFCs). We also need
an applicability statement.

The status of the IS-IS / IDRP&BGP interaction is unknown. This should be
published as an Internet Draft and/or RFC.

Changes to I.IS-IS since last time (mostly clarifications resulting from
testing): All changes will interoperate with implementations based on the
existing spec. Some of these changes result in changes to the base ISO
spec: These will be sent to ISO, and will also become a normative part of
the I.IS-IS spec.

Precedence of ES-IS information versus IS-IS information: There could be
cases where there is an adjacency learned from ES-IS (corresponding to a
subnet with a high metric), and also a lower metric path learned from
IS-IS. In this case, you should  go with the lowest cost path. Some
implementations do this the other way currently, and/or provide a knob to
allow folks to control this.

The same issue occurs with IP routes. The same solution should be used.
This is already described in the I.IS-IS specification (although there is
the option of having a configuration option to override the metrics).

There are a couple of changes to the configuration options in the MIB: (i)
Reachable address prefixes to include the NET; (ii) ES polling timers
(ES-IS Configuration Timer option) -- allow user to turn off the rapid
polling option (ageing only).

Padding in IS-IS Hellos can be turned off (this is particularly important
for low bandwidth and high cost links). Also, you must not discard a Hello
packet just because the neighbor uses a different pad size.

Uses of area addresses: The relationship between the configured area
addresses and the computed area addresses is confusing to some implementors
and users. You should use the configured area addresses in computing end
system adjacencies. This should be a transient problem and/or gross
configuration error, and therefore should not happen much. We discussed
whether to use the union of computed and configured, or just computed. It
seems that using configured is broken. We will resolve this on the mailing
list.

Routing through overloaded ISs. Implementations are not currently
consistent (ie, the existing text is in fact correct, but has been
mis-interpreted. The text therefore has been clarified to emphasize that
the goal is to be able to reach the overloaded IS, but not other end
systems "after" the overloaded IS.

Designated IS TOS Reporting: DISs currently only report the TOSs that they
themselves support when generating pseudonode LSPs (reporting end nodes).
They should report all TOSs. Clarifying text has been added (some existing
implementations have already fixed this).

Interfaces to subnets where you are not running IS-IS: Clearly you want to
be able to reach such subnets (even if you don't run IS-IS on it).
Clarifying text has been added.

Given fixes to the one minor issue discussed above, the overwhelming
working group consensus is that I.IS-IS is then ready to go to draft
standard (and the MIB to proposed standard). Any significant additional
features would then be added to a future ("version 2") of the standard.