The MPLS WG Archive

[Kireeti Kompella <kireeti@juniper.net>]

> I believe there is a reason.  And I've yet to be convinced otherwise by
> anyone.  The way I see it, peer model = single control plane.  Single
> control plane = common addressing, signalling, & routing.  Common addressing
> = only possible if the OTN has a single client.

I don't agree with this last statement: why is this the case?

I see a perfectly viable network with a single control plane for
"internal use" AND a UNI interface for other "external" clients.
Partition the lambdas in the optical network into "for internal use"
and "for external use", if that's what is required -- use some policy
mechanism, or do it by configuration.  If the "external clients" are
IP-based, just don't share your addresses with them.  In today's IP
networks, IGP addresses are not shared across external BGP sessions.
Or use IGP areas to isolate addresses.

However, the point of GMPLS is NOT a peer model.  It is an IP-based
control plane, leveraging existing work, and the flexibility of a
peer or overlay or hybrid model.

Kireeti.