El 25/07/2011 19:20, JP Vasseur escribió:
So let the WG decide which one of the methods is most appropriate
Thanks.
Dear PCErs,
Just my (subjective) opinion:
IIUC, the motivation behind this draft is the lack of upstream TE
attributes in inter-AS links (due to the lack of proxy), which is
necessary to compute bidirectional LSPs.
For this, the draft extended BRPC, proposing that the PCEP response
includes links that respect the constraints from of the request, for the
reverse direction, sent by the downstream PCE. Per se, I am not against
augmented procedures where applicable.
However, I believe that the original/main requirement not only applies
to BRPC but also when a Path is "segmented" (i.e. a (generic) client
requests a path from an entry node to an exit node, even without the
VSPT to later compose a path) or with H-PCE. I tend to think that (even
if sometimes the contrary seems easier) TED management (including
"augmented TED" with knowledge of border nodes, areas, and inter-domain
links) and Path Computation functions should be kept, as much as
possible, decoupled. Since the PCE architecture does not specify how a
PCE obtains the TED, I would prefer to maintain both concepts separated,
and embedding the links along with the VSPT does not seem to fit, much
alike alternative solutions which could be proposed where PCEs could
inform each other of TE attributes of AS links (embedded in
notifications). After consideration, this may not be a good idea.
I am aware that my comment is of little help since, even if I prefer
that the scope of responsibility of PCE(i) starts in its entry border
nodes and PCE(i-1) trims the VSPT based first on the Inter-As links and
second on its domain, I am not solving the original problem/requirement
other than "PCE(i) must also know the TE attributes of the upstream
link" but I don't know / consider how
Thanks and best regards
Ramon
_______________________________________________
Pce mailing list
Pce@ietf.org
https://www.ietf.org/mailman/listinfo/pce