Re: FW: Questions on draft-ietf-mpls-bundle-00

[Yakov Rekhter <yakov@juniper.net>]

Ed,

> draft-ietf-mpls-bundle-01.txt clearly states that only TLVs of types 1
> (IPv4), 2(IPv6) or 3(IF_INDEX) are to be used in the IF_ID object (when
> dealing with component links).  

correct.

> However, types 4 & 5 (the COMPONENT_IF_*
> types) are still defined in the new GMPLS draft
> (draft-ietf-mpls-generalized-signaling-07.txt).

that is correct too.

> Should they have been retired from this latest revision of the generalized
> signaling draft?

I wouldn't argue against this :-)

> Also, draft-ietf-mpls-generalized-rsvp-te-06.txt specifies that, for
> component links, it is still possible to specify distinct downstream and
> upstream component data channels:
> 
>    ...In the special case where a
>    bidirectional LSP that traverses a bundled link, it is possible to
>    specify a downstream data channel that differs from the upstream data
>    channel. 
> 
> If one shouldn't use the COMPONENT_IF_DOWNSTREAM and COMPONENT_IF_UPSTREAM
> TLVs, then presumably one would have two TLVs in the IF_ID hop object.  Is
> there a convention which says which TLV refers to the downstream and which
> the upstream direction?

If there are sufficient *practical* justification to support the
special case (downstream channel different from the upstream
channel), then we would add text to the link bundling draft to
cover this case by using COMPONENT_IF_DOWNSTREAM and COMPONENT_IF_UPSTREAM
TLVs.

Yakov.