RIP-2 Addresses Links and MIB problems

[Jeffrey C Honig <[e-mail]>]

What's the status of RIP-2 on the protocol standards track?  When is
the next chance for changes to the documents?

I've been looking at the MIB to see what is involved in implement it
and noticed a couple problems.

First, interfaces are referenced in the MIB by their IP addresses, for
point-to-point interfaces I assume this means their local addresses.
A frequently used capability of Berkeley Unix is to have the local
address of one or more point-to-point interfaces share an address with
a non point-to-point interface (i.e. an Ethernet).  This is similar to
in the Router Requirements draft as a the "router-id" method of
dealing with "unnumbered" links.

As the RIP MIB is specified, I would need to report multiple
interfaces with the same address, which SNMP is not capable of dealing
with.

One option, which I use for all interface references in gated, is to
represent point-to-point interfaces by their destination address,
since that almost certain to be unique.  The other option I see is to
represent these interfaces as unnumbered and add an interface Index
field in the manner of OSPF.

As to addressless interfaces in general, is there a specific reason
they are not supported in RIP-2 (and RIP-1).  It is not necessary to
know the address of your peer to send packets, that can be done by
sending to the RIP multicast address (or all ones broadcast), you'll
learn the address of your peer as a packet is received.


Second, as I understand Domains (and I think much of the blame for
them lies on me), it is possible to run multiple RIP Domains on an
interface.  This would allow reaching a different subset of routers on
the interface with each Domain.  The MIB does not allow for this, it
should have rip2IfConfDomain as an INDEX of rip2IfConfTable.  And
shouldn't the rip2IfStatTable have a second index of rip2IfConfDomain
also to provide stats per domain?


Third, rip2IfConfReceive implies the ability to selectively listen to
version 1 packes, version 2 packets or both on a per interface (and
hopefully per Domain) basis.  This capability is not mentioned in the
RIP2 spec and I hate it when a MIB adds functionality not mentioned in
the protocol spec.  Can the protocol spec be updated to reflect this?

Jeff