Re: [mpls] Mixing ICC and Global-IDs in MPLS-TP Identifiers?

["Adrian Farrel" <[e-mail]>]

Hi Huub,

Converging your two emails...

> > So, I may have understood the import of those figures (specifically
>  > figure 3 in draft-palanivelan-bfd-v2-gr-11.txt)
> 
> Just to avoid confusion I was refering to the figures in
> draft-tsb-mpls-tp-ach-ptn-00.txt and looking at the text below
> you may be refering to that draft too.

Yeah! Sorry about that. Sloppy cut and paste. So many I-Ds and so little time.

> > but I read that to  mean that, in the case of mismatched OAM types,
>  > one of the OAM types (in the figure, the PSN type is indicated)
>  > would be run end-to-end, and the local OAM types would be used
>  > within the networks according to what they supported.
> 
> Indeed.
> 
> > Now, noting very clearly that OAM types have nothing whatsoever to
>  > do with MPLS-TP Identifiers, you seem to be suggesting that the
>  > same approach holds.
> 
> The OAM types are independent of the used identifiers.
> 
> > That would mean that (to paraphrase Figure 3) network A might use
>  > one form of identifier and network B might use another form of
> >identifier,
> 
> yes.
> 
> > but the end-to-end identifiers would be of one form only.
> 
> Indeed, and the identifier that is used end-to-end can be either
> form A or form B. The actual form has to be agreed between the
> operators of network A and B.

| This can mean that in the direction A --> B one form can be used
| (i.e. the form that operator A uses in his whole network) and
| that in direction B --> A of the same LSP/PW form B can be used
| (the form that operator B uses in his network).
| At the sink MEP the identifier has to be verified, the form can
| be independent of the form used in the network. Operators A and B
| inform each other about the "value" of the expected identifier.

I'm not clear about this.
It is true that this *could* be done, but it also seems that it would be a bit odd to mix and match like this for a bidirectional LSP. However, I can see that for individual LSPs the behavior could be one of:
- always be configured so "A runs over B"
or
- always depend on the initiating network
or
- one mechanism always runs over the other

My interpretation of the OAM discussion was that "once the IETF OAM mechanism was standardised, all edge nodes would need to support it" so that we would have to have a case where the third option was the only one available. 

Now, regardless of all that, it seems that we have reached agreement that, while the end points might need to support both identifier formats, there would never be a case where mixed identifier formats were used.

Cheers,
Adrian

> > Have I misunderstood you?
> 
> I don't think so.