Your opinion: Using the RDB to make TE decisions

["Adrian Farrel" <adrian@olddog.co.uk>]

Hi,

In my review of draft-ietf-ccamp-inter-domain-pd-path-comp-00.txt (q.v.) I
found that the authors are proposing that when an ERO next hop cannot be
resolved using the TED, it MAY be resolved using the routing database.

That is, in the absence of known TE links all the way to the next hop, the
path computation component MAY determine some LSR to which it does have
TED connectivity and which it believes has IP reachability to the next
hop.

The proposed use for this is to resolve TE connectivity to out of domain
(e.g. out of AS) next hops. In particular to find the best ASBR to use to
get to the destination.

To be fair, the authors do not claim that their technique will produce the
most optimal route. Nor do they claim that it will always find a route
without using crankback.

1. I have asked them to tighten their language so that the use of
   the RDB is not allowed for TE route computation *within*
   a domain.
2. What does the WG think about this proposed use of the RDB
   for inter-domain route resolution within GMPLS networks.
   Note that the next hop is a TE address:
   a. Where this is a TE router ID the process may work fine.
   b. Where this is a TE link address and not present in the RDB
        no harm is done.
   c. The TE link addresses MIGHT re-use addresses from the
      IP control plane but identify different resources. This would
      be a mess.

Your opinion is needed.

Thanks,
Adrian