Michael Thomas wrote:
>
> Brian E Carpenter writes:
> > Michael Thomas wrote:
> > >
> > > Brian E Carpenter writes:
> > > > Excellent summary. We originally chose a 6 bit diffserv field partly because
> > > > it was available in both IPv4 and IPv6, and partly because it allows for
> > > > very efficient classification in *core* routers, with the more demanding
> > > > multi-field classification being left to border routers.
> > > >
> > > > The question before the house (in the end that means both ipng and diffserv)
> > > > is whether the added complexity of adding the flow label to the diffserv
> > > > model is justified by the gain in expressiveness. It doesn't do anything
> > > > for the trust model.
> > >
> > > I haven't heard about any imminent shortage of DSCP's.
> > > Indeed, it seems that there's only a small handful
> > > (2-6) that I've ever heard people contemplating. Maybe
> > > I just don't travel in the right circles...
> >
> > We have deliberately been *very* conservative about defining standard PHBs, since
> > we want to be very sure about what we standardise. But there are a potentially
> > infinite number of local-use PHBs. That is why the DSCP value is mappable, and why
> > the PHB ID was defined, so that local-use PHBs can be registered with IANA and
> > signalled. The idea here is to stretch the semantics of the PHB ID just a little.
>
> Right. Even if the number of PHB's grows beyond
> 6-8 bits (which seems pretty unlikely to me), in
> order to be a real problem you would have to posit
> a network/domain which actually wanted to use more
> than could be currently mapped into the DSCP. I've
> been working on a lot of voip stuff in the last
> few years and the number of DSCP's I've heard has
> been in the range of about 2 to 5 or 6 that I
> recall, with higher end of the range seeming
> pretty extravagant.
>
> > Normal semantics of a PHB ID:
> >
> > "This is local-use PHB number 379"
> >
> > Stretched semantics:
> >
> > "This is local-use PHB number 379 and according to our SLA, that gets classified
> > as real-time traffic needing at least a 10 Mbit/s rate"
>
> Ah. This strikes me as trying to overload the
> semantics of DiffServ.
Yes. Pretty shameless of me as diffserv co-chair :-)
> Another way to go about
> this is to use signaled diffserv (either the
> COPS-PR or RSVP varieties) with policers at the
> proper edges of the network. That way, the traffic
> classes (ie the actual PHB) stays small.
True. But signalling in the flow label is a low overhead approach
(although of course there needs to be an out of band way of
creating the SLA).
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]
--------------------------------------------------------------------
