The MPLS WG Archive[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index][Thread Index][Author Index][Subject Index] BNF for GMPLS path message (was Re: draft-gray-mpls-rsvp-oif- uni-ext)
At 06:24 PM 9/18/00, Eric Gray wrote: >Lou, > > Now, if you just put the SUGESTED_LABEL >back together with the LABEL_SET, you will have >what we had originally. yes, but they are both needed as they serve different purposes. > :-) > > What application do you have in mind for >using both a SUGESTED_LABEL and a LABEL_SET? LABEL_SET limits the label space that can be used by the downstream node. If none of the labels are acceptable, the downstream node responds with an error. See section 3.5 of draft-ashwood-generalized-mpls-signaling-00.txt for a description where this is anticipated to be useful. SUGGEST_LABEL is just that, a suggestion. If the downstream node doesn't like the suggestion, it can choose whichever label it considers to be available. The purpose of the suggested label is to reduce hardware setup time in the upstream node. See section 3.4 of the draft for more info. >Semantically, there's nothing non-sensical about >having both, but the suggested applications for >each are fairly diverse. One is for latency, the other is for limited switching capabilities... Lou >-- >Eric Gray > > > -----Original Message----- > > From: Lou Berger [mailto:lberger@labn.net] > > Sent: Monday, September 18, 2000 3:01 PM > > To: Adrian Farrel > > Cc: Lou Berger; egray@zaffire.com; mpls@UU.NET > > Subject: RE: BNF for GMPLS path message (was Re: > > draft-gray-mpls-rsvp-oif- uni-ext) > > > > > > At 05:20 PM 9/18/00, Adrian Farrel wrote: > > >Thanks Lou, > > >That makes some sense. > > > > > >For <LABEL_REQUEST> should I read > > ><LABEL_REQUEST> | <GENERALIZED_LABEL_REQUEST> or > > >are some of the fields of GLR sufficiently > > >specific to the sender for that object to need to > > >be in the sender descriptor? > > > > It's the same class so you can only have one. I think the > > generalized one > > c-type = 5 can always be used. > > > > >Is it also right to add <LABEL_SET> to > > ><sender descriptor> ? > > > > Woops, not that got missed. > > > > I think location is arguable either way, but I believe > > outside the sender > > description is preferable (for other message processing). My > > suggestion is: > > > > <Path Message> ::= <Common Header> [ <INTEGRITY> ] > > <SESSION> <RSVP_HOP> > > <TIME_VALUES> > > [ <EXPLICIT_ROUTE> ] > > <LABEL_REQUEST> > > <LABEL_SET> > > [ <SESSION_ATTRIBUTE> ] > > [ <POLICY_DATA> ... ] > > <sender descriptor> > > Lou > > > > > > >Adrian > > > > > > >From: Lou Berger [mailto:lberger@labn.net] > > > >Sent: Monday, September 18, 2000 9:53 PM > > > > > > > >Adrian, > > > > My understanding of the BNF for Path based on the > > generalized > > > >draft is: > > > > > > > > > > > > <Path Message> ::= <Common Header> [ <INTEGRITY> ] > > > > <SESSION> <RSVP_HOP> > > > > <TIME_VALUES> > > > > [ <EXPLICIT_ROUTE> ] > > > > <LABEL_REQUEST> > > > > [ <SESSION_ATTRIBUTE> ] > > > > [ <POLICY_DATA> ... ] > > > > <sender descriptor> > > > > > > > ><sender descriptor> ::= <SENDER_TEMPLATE> <SENDER_TSPEC> > > > > [<ADSPEC>] [<RECORD_ROUTE>] > > > > [ <SUGGESTED_LABEL> ] > > > > [ <UPSTREAM_LABEL> ] > > > > > > > >It was in the draft and was removed due to lack of parity > > between the > > > >protocols. Obviously it needs to be in the eventual RSVP > > related spec. > > > > > > > >Lou > > > > > > > >At 04:26 PM 9/18/00, Adrian Farrel wrote: > > > >>Hi Eric, > > > >> > > > >>Thanks for this draft, it pulls things together nicely and > > > >>shows how simply UNI can be performed with existing protocols. > > > >> > > > >>I have just a couple of questions. > > > >> > > > >>Is the Propagation Delay object wholly new? Have you any > > > >>plans for what it looks like? > > > >> > > > >>In sections 3.4 and 5.4 (which need renumbering :-) you use > > > >>a Notify to expedite the acknowledgement of a Tear if there > > > >>is no suitable message on which to piggy-back the Ack. Why > > > >>did you choose this rather than an Ack message? The Ack > > > >>message exists for exactly this purpose, while the Notify > > > >>requires an Error_Spec. > > > >> > > > >>I believe you may have cut and pasted a typo from > > > >>draft-ietf-mpls-rsvp-lsp-tunnel-06.txt. George has fixed > > > >>this in v7. TIME_VALUES should be mandatory on Path and Resv > > > >> > > > >>Lastly, I'm not sure that the format you have given for the > > > >>Path is quite correct with regard to the GLR, LS and SL > > > >>objects. Since my version of GMPLS doesn't have anything > > > >>specific to say on the subject I can only piece it together > > > >>from the text... > > > >>- I don't think you can have Generalized Label Request > > > >> and suggested label on the same Path message. > > > >>- I think Label Set supplements both Generalized Label > > > >> Request and Suggested Label. > > > >> > > > >>This (and the previous point) would give > > > >> > > > >> <Path Message> ::= <Common Header> [ <INTEGRITY> ] > > > >> <SESSION> <RSVP_HOP> > > > >> <TIME_VALUES> > > > >> [ <EXPLICIT ROUTE> ] > > > >> <GENERALIZED LABEL_REQUEST> | > > <SUGGESTED LABEL> > > > >> [ <LABEL SET> ] > > > >> [ <UPSTREAM LABEL> ] > > > >> [ <SESSION_ATTRIBUTE> ] > > > >> [ <POLICY_DATA> ... ] > > > >> <sender descriptor> > > > >> > > > >>Regards, > > > >>Adrian > > > >>-- > > > >>Adrian Farrel mailto:af@datcon.co.uk > > > >>Network Convergence Group > > > >>Data Connection Ltd., Chester, UK > > > >>http://www.datcon.co.uk/ > > > >>Tel: +44 (0) 1244 313440 Fax: +44 (0) 1244 312422 > > > > > >
|
|