Steve Blake wrote:
>
> Brian Carpenter wrote:
>
> > Thomas Narten wrote:
> > ...
> > > In my view, deciding on the right useage of the bits requires first
> > > understanding the problem that needs to be solved.
> >
> > I think that from the QOS viewpoint, we do understand the problem very well,
> > hence the proposal to add a diffserv usage to the existing intserv usage.'
>
> When I think about the most popular usage of protocol/port classification
> for QoS (clamping down Napster traffic on university LANs), I don't see
> how the new flow label proposal will solve that problem.
I hope not. It wasn't designed to. At the risk of repeating myself,
the idea is to partially alleviate the problem of classifying
encrypted packets.
> Protocol/port classification is only a "requirement" of Diffserv because
> we haven't solved the problem of enabling the host to put meaningful
> values in the DSCP. I would prefer to try to solve that problem.
That is really a small matter of programming. AIX has been able to set
DSCPs at source for a couple of years, and so has Linux and one or
two other o/s. The issue is getting QOS policy managers into
the hosts, and optionally adding code to applications to interact with
the policy manager and set the DSCP on the socket. IMHO there is
no IETF issue here.
> > The intserv usage is documented and somewhat implemented, and I don't
> > think we can hide behind the weasel words in RFC 2460 any longer.
>
> I think the requirement that the flow label be a PRN could be relaxed:
> as long as it is unique per-host it should suffice for Intserv and other
> flow-caching usages.
Probably. I don't really remember the argument for PRN but it may have been
gut feeling about hashing efficiency.
>
> > I don't pretend to understand the routing handle usage (but then,
> > I'm of the school of thought that doesn't understand why anyone would
> > want MPLS either).
>
> Because some folks want Frame Relay NG, for a variety of reasons, including
> multi-protocol support. Which is why trying to fit label switching into
> IPv6 is completely besides the point.
Fully agree. When we rejected CATNIP in 1994, we rejected this.
Brian
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
