draft-shiomoto-ccamp-lsp-hierarchy-bis-00.txt

["Don Fedyk" <dwfedyk@nortel.com>]

Hi

Thanks for the presentation yesterday. A thought occurred to me about
the numbered versus unnumbered issue in your draft. (Note I will not
refer to the multiple instances of the control plane since as discussed
in the meeting I think this should be out of scope).   

What I like about the unnumbered identifiers are they are hierarchical
under the GMPLS node ID. I would suggest that even when an interface is
numbered it could have an unnumbered identifiers as well and the
numbered interface not be assumed to be advertised but assumed instead
to be an identifier. It is in my opinion a special case when the
numbered interface is out of the same address plan as the GMPLS control
plane.  Generally we should not assume this is true and separating the
local identifier context from the GMPLS control addressing has
advantages.  (already illustrated by your draft). 

If we treated all links this way they could have different identifiers
for the purpose of identification and signaling in the link identifier
space (e.g. IP, MAC, NSAP etc) and uniformly use unnumbered identifiers
for all routing and TE advertisement.  I realize this is not the way
that unnumbered links came about but this might be the more
"generalized" way to address this.

Did you consider this option?

Regards,
Don