Re: BGP-4+

[Dennis Ferguson <[e-mail]>]

Yakov,

> This is certainly one possibility. Another possibility is to allow
> MP_REACH_NLRI and MP_UNREACH_NLRI to carry NLRI for multiple address
> families (this would require minor changes to the proposed encoding).

This would work too.  Now that I've though about it I think I might have
a slight preference for doing it your way.  The tradeoff seems to be that
you have to spend a few more bytes of encoding to do it your way, but
you get to keep all attributes-that-are-really-routes together in one spot.
I think the latter is probably a feature.

The only other issue I have relates to protocol-specific usage.  I think
there probably needs to be a paragraph or so on IPv6-specific gorp written
down somewhere to ensure working, interoperable implementations.  The one
example I can think of off-hand is that of IPv6 next hop addresses.  Most
routing protocols for IPv6 specify the exclusive use of link-local addresses
for next hops, as the interface addresses least likely to change over the
life of a routing protocol session, but this is probably not right for BGP
since link-local addresses are not likely to be well-received when
readvertised via IBGP.

Is it intended that protocol-specific (or, at least, IPv6-specific) issues
are to be documented elsewhere (or anywhere)?  A multiprotocol specification
by itself seems insufficient.

Dennis