The MPLS WG Archive

[jeremy.de_clercq@alcatel.be]

Dave, all,

> If we go on the assertion that MPLS should have had a PID (as opposed
> to current practice of binding protocols to labels) then we need to
> scope what we are really trying to achieve....
> 
> 1) Make the payload self describing to end points.
> 2) Multiplex payloads over a single LSP.
> 3) Make the payload self-describing to intermediate systems.
> 
> At the moment, 3 appears to be the exclusive motivation for this but
> has implications in 1 & 2 as we already have mechanisms for achieving
> those goals today (indirectly via signalling (1) & label stacking
> (2)).

If MPLS should have a PID, is it something that would be mandatory on
every LSP, or is the presence of this to be signaled (e.g. using a new
'PW type' in draft-ietf-pwe3-control-protocol (e.g. "type
Multi-Protocol")) ? I think that if MPLS PID is defined, the use of it
should be optional and signaled. So not "as opposed to binding protocols
to labels", but "as an optional alternative to binding protocols to
labels".

Applications for motivation 1) are described in
draft-moreels-multiproto-mpls-00.txt and in
draft-sajassi-l2vpn-interworking-01.txt.

thanks,
Jeremy.