Internet Info 1994 March/Internet Info CD-ROM (Walnut Creek) (March 1994).iso/inet/ietf/isis/isis-minutes-91jul.txt

Text File | 1993-02-17 | 8KB | 192 lines

CURRENT_MEETING_REPORT_


Reported by Ross Callon/DEC

IS-IS Minutes

The IS-IS Working Group met at the IETF meeting in Atlanta, Georgia.
There were two topics of discussion:  A brief overview of the status of
the IS-IS spec (led by Ross Callon), and a presentation and longer
discussion of the SNMP MIB for IS-IS (led by Chris Gunner).


  1. Status of IS-IS
     Ross reported that the OSI IS-IS Intra-Domain routing protocol (ISO
     DIS 10589) has completed the Draft International Standard (DIS)
     ballot, and all ballot comments were successfully resolved at a
     recent ISO meeting.  This implies that the ISO IS-IS will be
     progressing to final International Standard state relatively
     quickly.  This, in combination with the completion of a couple of
     Integrated IS-IS implementations means that it is a good time to
     start think about issuing an update to RFC 1195.

     Ross then gave a quck overview of some minor changes that would be
     involved:


       o Reference to ISO standard
         RFC 1195 reference the DP version of ISO IS-IS. This clearly
         needs to be updated to reference the final International
         Standard version, when available.  This would also imply that
         Annex B (Encoding of Sequence Number Packets) can be removed.
         It turns out that we were either lucky or good, and the
         sequence number format in the current ISO document is
         compatible with Annex B.

       o RIP (or other external routes) at level 1
         Currently the spec says that this is not allowed.  There are
         good technical reasons why we don't want fully general external
         connections at both level 1 and level 2.  However, there may be
         many cases where we have a small RIP ``island'' which is only
         reachable via a level 1 area.  For example, this is very likely
         to occur during transition from a RIP routing domain to an
         Integrated IS-IS routing domain.  No technical change is
         needed, but the document should be upgraded editorially to
         specify that this is permissible.

       o Default IP route at level 1.
         There will be some cases where level 1 routing is IP-capable
         (using Integrated IS-IS) but level 2 routing is not (such as
         using OSI-only IS-IS at level 2, or possibly during phase 4 to
         phase 5 DECnet(TM) transition).  In this case, there needs to
         be a way for level 1 routers to know where to send traffic

                                   1





       destined to outside of the area (for example, one single level
       2 router might be running RIP with external routers).  The
       solution to this is to allow IP Default Route (subnet mask of
       all 0's) at level 1, and to specify that for level 1 only
       routers which see the default route advertised in level 1 LSPs,
       this takes precedence over forwarding traffic to level 2
       routers.

     o Compatibility with earlier versions of IS-IS
       There should be a ``for information only'' annex which
       specifies the differences between RFC 1195, and the updated
       RFC. This will also specify how to ensure interoperability
       between old and new routers.

     o IS-IS / BGP interaction
       Yakov Rekhter brought up the issue of interaction between IS-IS
       and BGP. Ross and Yakov will work on this issue off-line, and
       report results back to the Working Group.

     o Encoding of Authentication Field
       Someone brought up the issue that RFC 1195 and DIS 10589 both
       have an authentication field, in which the encoding and use is
       identical but the code value is different.  The Working Group
       agreed that this was an unnecessary redundancy, and that we
       should use the value from 10589.

     o Ships in the Night Operation
       RFC 1195 currently has sufficient functionality to allow
       operating two instances of IS-IS in ``Ships in the Night'' mode
       -- one instance would be for IP-only routing, and one for
       OSI-only routing.  However, just how to do this is not written
       down anywhere.  It was agreed that this should be writted down,
       with the approach ``you don't have to be capable to run two
       instances of IS-IS, but if you do run two instances then this
       is how you do it''.  Generally, you demultiplex on the
       ``Protocols Supported'' field, and optionally may use
       authentication to protect against accidental merging of the two
       logical routing domains by a mis-configured router.


2. MIB for IS-IS
   Chris Gunner then gave a detailed presentation of the proposed MIB
   for IS-IS. This MIB allows management of Integrated IS-IS
   (including full management of both ISO 10589 and RFC 1195) using
   SNMP. This is based on the GDMO (i.e., ISO format network
   management information) contained in DIS 10589, with additional
   objects added for management of RFC 1195.

   The recent progression of 10589 in ISO will result in some changes
   to the GDMO in 10589.  Chris will need to produce an update of the
   MIB in order to maintain alignment with the ISO document.



                                 2





     There was a discussion of the size of the MIB. In particular, there
     are situations where several similar things are in different tables
     For example, different sorts of circuits currently are managed
     using different tables.  There is substantial overlap between these
     different tables.  The alternative is to have one type of table for
     all circuits, with some fields not always used.  This implies
     slightly more bits will be transmitted on the wire, but allows a
     smaller MIB and less software code (e.g., data structures are
     simpler).  The Working Group agreed that the latter approach was
     preferable, at least in those cases where the overlap is relatively
     large.

     The group agreed that the MIB should permit multiple instances of
     Integrated IS-IS and/or IS-IS to be managed in a system.  This
     means turning single instance objects in groups into table objects.
     The group also agreed that all such table entries should be capable
     of creation and deletion to mirror the creation and deletion
     capabilities of the DIS 10589 managed objects to which they are
     equivalent.

  3. Other Issues
     Yakov Rekhter pointed out that the ISO GDMO of IS-IS does not allow
     measurement of routes coming from external protocols to IS-IS.
     Chris and Ross agreed to bring up this issue with the folks working
     on the ISO specification.

     Outside of the Working Group, a couple of folks brought up the
     issue of how to handle the ``3rd party router'' case (a single
     routing domain having several routers on a broadcase or
     general-topology network with only one router running BGP). Ross
     will write up a proposal on how to deal with this and discuss it
     within the Working Group.


Attendees

Nagaraj Arunkumar        [e-mail]
William Barns            [e-mail]
Scott Barvick            [e-mail]
William Biagi            [e-mail]
John Burruss             [e-mail]
Ross Callon              [e-mail]
Dino Farinacci           [e-mail]
Dennis Ferguson          [e-mail]
Robert Griffioen
Chris Gunner             [e-mail]
Robert Hagens            [e-mail]
Susan Hares              [e-mail]
Manu Kaycee              [e-mail]
Paulina Knibbe           [e-mail]
Dale Land                [e-mail]
Chao-Yu Liang
Shane MacPhillamy        [e-mail]

                                   3





Bill Manning             [e-mail]
Dennis Morris            [e-mail]
Yakov Rekhter            [e-mail]
Mike Truskowski          [e-mail]
Rick Wilder              [e-mail]
L. Michele Wright        uncng![e-mail]



                                   4