Thomas,
This is or may be a sensitive topic for many.
The reality is that, at least a majority, if not all, of the original
goals of the Flow ID field
are achieved by the MPLS shim header. Remember the discussions of many
years ago, of where to place the flow ID relative to the source and
destination addresses? Well that was also answered by the MPLS header.
Furthermore, The MPLS label distribution protocols - RSVP is one of
them - are the mechanisms used to assign and distribute labels. IPng WG
never really tackled this area, and in reality there is no protocol that
defines a consistent IPv6 flow ID allocation, and distribution
that I am aware of.
Furthermore, the ability to build hierarchies, the ability to use
directly fast ATM, and FR
hardware forwarding engines, and the better traffic engineering
capabilities, give a considerable
edge to MPLS, over the IPv6 flow ID.
Considering these, and perhaps more, I think the IPv6 flow ID field is
obsolete, and could be
redefined as "reserved", or renamed, and given a more useful role than
it has now.
I may put out a draft on this topic, to generate or add to a discussion
point on the agenda, if one already exists.
Alex
Thomas Narten wrote:
>
> Metzler,
>
> > Within RFC2460 the usage of the flow label field is not
> > fully described and for me it is not clear if anything
> > mentioned in the Appendix is part of the standard or
> > only for information. Therefore I have the following
> > questions:
>
> The bottom line is that as of this time, no one has proposed how they
> would like to use the Flow Label and led the WG to consensus on their
> approach. Hence, it really is unspecified in the sense that its
> useage/definition remains to be defined.
>
> If you have a proposal on how it could/should be used, this WG is
> certainly the place to make that proposal, and I would encourage you
> to do so.
>
> I'll note that there have been discussions in the past where arguments
> have been made that the flow label should not be modified by any
> routers (i.e., it has end-to-end semantics that need to respected)
> while other folks have argued that it might be good to allow it to be
> modified by routers (e.g., in support of something like MPLS). The
> door has been left open for either of those approaches. Specifically,
> the Flow Label is considered a mutable field from the perspective of
> IPsec, thus IPsec won't break if the field is modified by routers.
>
> Again, what is needed is a concrete proposal that answers the types of
> questions you raised.
>
> Thomas
> --------------------------------------------------------------------
> IETF IPng Working Group Mailing List
> IPng Home Page: http://playground.sun.com/ipng
> FTP archive: ftp://playground.sun.com/pub/ipng
> Direct all administrative requests to [EMAIL PROTECTED]
> --------------------------------------------------------------------
S/MIME Cryptographic Signature
