The MPLS WG Archive

[Eric Gray <EGray@zaffire.com>]

Kireeti,

	Please see embedded comment.

> -----Original Message-----
> From: Kireeti Kompella [mailto:kireeti@juniper.net]
> Sent: Wednesday, August 16, 2000 11:50 AM
> To: akullber@netplane.com; EGray@zaffire.com; kireeti@juniper.net
> Cc: mpls@UU.NET
> Subject: RE: generalized MPLS signaling question
> 
> 
> Hi Eric,
> 
> > 	But this fits snugly under the heading of a "neat
> > trick" and it would be clearer and less likely to cause
> > trouble down the road if some mechanism were used to
> > specify an LSP explicitly.
> 
> Actually, if you really want to choose ("force") the path,
> you'd use strict hops.  If LSPs/interfaces are uniquely
> numbered, there isn't (shouldn't be!) any ambiguity.

I agree that this should work.  I also like the
approach of not having a different answer for
MPLS and other link types.

What I can't seem to find is where it is made
clear that an LSR would necessarily find this 
one unique IP address route.  As I said earlier, 
it makes sense, but I don't think we can get 
away with not specifying something explicitly 
because it makes sense.  Lots of things make 
sense.

I looked through the following drafts: 	

	draft-kompella-mpls-optical-00.txt
	draft-kompella-mpls-bundle-02.txt
	draft-kompella-lsp-hierarchy-00.txt
	draft-kompella-mpls-unnum-00.txt

I was looking for something that said something
like:

"When processing an ERO, if a forwarding adjacency
 has an IP address that exactly matches the address
 in the ERO sub-object (in the case of RSVP-TE), 
 the LSP being established is expected to traverse
 the forwarding adjacency LSP using label stacking.

"If the ERO's next sub-object is an interface index
 sub-object and there is a corresponding unnumbered
 forwarding adjacency with a matching interface index,
 the LSP being established is expected to traverse
 the forwarding adjacency LSP using label stacking."

If something like this exists already, can you point
it out?

In fairness, you could argue that this is implied if
the forwarding adjacency is advertised into an IGP.
But advertising a forwarding adjacency into a specific
IGP is not mandated in any of the drafts noted above.
Yet, I assume that - if the LSP initiator somehow got
a hold of the IP address for such a FA-LSP - we would
still want the corresponding LSP used.  If not, then
this is not clear either.

Thanks!

> 
> Kireeti.
>