The MPLS WG Archive

[Eric Gray <eric.gray@sandburst.com>]

David,

    Maybe I missed something: I thought support for unnumbered links
needed to be added for GMPLS because there are problems with these
links in the general case.  I was not aware that there are problems with
unnumbered links in the Ethernet, PPP, ATM and Frame Relay cases or
in basic signaling of MPLS explicit routes.

    What exactly are the problems with using unnumbered links in the
specific cases defined for MPLS?

You wrote:

> Yakov Rekhter wrote:
> >
> >> 4: Is there any way for an ingress router to compute an ERO with
> >>    unnumbered links in it if GMPLS is not implemented?  It appears
> >>    that support for unnumbered links was deleted from the TE
> >>    versions of OSPF and ISIS, and moved into the GMPLS versions of
> >>    those protocols, making the necessary information for ERO
> >>    computation unavailable in an MPLS environment.
> >
> > You certainly don't need to implement all the GMPLS routing and
> > signaling extensions in order to support unnumbered interfaces.
>
> Understood.  But if an ERO is to be auto-generated, and not manually
> entered, the box doing the computation needs to know the topology.
> Without the unnumbered extensions in OSPF/ISIS, any topology derived
> from routing will not distinguish multiple parallel unnumbered links.
> This will make it impossible to compute an ERO that uses the new ERO
> subobject to specify one such link.
>
> It wouldn't be very nice to require all of GMPLS in order to compute an
> ERO that specifies unnumbered links.
>
> I hope my explanation of this problem is clear.
>
> -- David

--
Eric Gray (mailto:eric.gray@sandburst.com)
http://www.mindspring.com/~ewgray