Gents, I’m 100% with Les here, going into platform/asic specifics within this document would inevitably create ambiguity.
Cheers, Jeff On Oct 2, 2018, 11:20 AM -0700, Les Ginsberg (ginsberg) <ginsb...@cisco..com>, wrote: > Bruno – > > Trimming the thread… > > [Les2:] Label imposition is meant to cover both the SWAP operation and the > PUSH operation. In the example you provided above where a label stack of “12” > is replaced by a label stack of “14,15” the number of labels “imposed” is 2. > [Bruno2] In that case, I definitely think that the discussion was useful and > that this point needs to be clarified in the document. > Whether you choose to call that (1 POP, 2 PUSH) or (1 SWAP, 1 PUSH) or > simply a SWAP isn’t relevant here (though it might matter to folks like the > RFC 3031 authors). > > With that ibn mind, here is proposed text: > > “Base MPLS Imposition MSD (BMI-MSD) signals the total number of MPLS > labels which can be imposed, including all service/transport/special > labels. Imposition includes swap and/or push operations. > > If the advertising router performs label imposition in the context of > the ingress interface, it is not possible to meaningfully advertise > per link values. In such a case only the Node MSD SHOULD be > advertised.” > > [Bruno2] Given that the term “imposition” does not seem to be defined within > the IETF, I would still favor a formal definition not using it. e.g. “BMI-MSD > advertises the ability to increase the depth of the label stack by BMI-MSD > labels”. > Alternatively, I’d propose the following rewording which seems clearer to me: > OLD: Imposition includes swap and/or push operations. > NEW: A swap operation counts as an imposition of one label; just like one > push operation. > > [Les3:] This gets into implementation specific issues that I would really > like to avoid. > For example, some implementations perform one and only one “operation”. > Conceptually that may involve a swap and a push – but from the internal > implementation POV it is simply one operation. And this may be true > regardless of how many labels are involved. Other implementations might > perform this in several discrete steps. The language we use here should not > imply anything about how many labels are associated with a specific operation. > > The term “increase” isn’t accurate because in the case of a swap there is no > increase, yet the label which is replaced is counted. > > https://tools.ietf.org/html/rfc3031#section-3.10 is relevant here. > > The term “imposition” is generic – and as Alvaro has pointed out is used in > RFC 4221. And the language proposed above does define the relationship > between “swap and push” and “imposition”. > > I appreciate your desire for clarity – and I am still open to new language – > but at this point I still think what I proposed is the most accurate. > > Les > > > > Thanks, > --Bruno >
_______________________________________________ Lsr mailing list Lsr@ietf.org https://www.ietf.org/mailman/listinfo/lsr
