Re: The problem of the framework draft

[Masataka Ohta <[e-mail]>]

Hal;

> Eric,
> 
> Good note (as always).

No, it's better neglected. But as you insist...

> >> >Why, do you think, we can't start modifying BGP, until we finish
> >> >modifying OSPF and RIP?
> >>
> >> I didn't say that we had to wait to modify interdomain protocols until we
> >> were finished with modifying intradomain protocols.
> >
> >What did you say, then?
> >
> 
> I said that we want to *start* with the intradomain protocols, not wait
> until all of them were finished before working on interdomain protocols!

We should better say that we want to *start* with the interdomain protocols,
not wait until a pile of intradomain protocols, which does not scale
nor interoperate with each other nor with scalable interdomain
protocols, are generated.

> >Then, why QoSR WG can say intra-domain should be tried first?

> We can suggest it and we also observe that this is what has already
> happened, witness the work on OSPF QoS routing solutions.

We can't suggest a thing which has already happened, becasue it
has already happened.

We can criticise QOSPF and discourage people repeat the same mistake,
instead.

> >> Part of the reasoning for starting with
> >> intradomain protocols is to gain further experience with QoSR methods in
> >> environments where experiments are easier to test and deploy.
> >
> >As I already stated, the easiest approach to test and deploy QoSR
> >is to use a protocol which does not distinguish inter/intra
> >domain.
> >
> 
> While part of your approach does not distinguish between interdomain and
> intradomain, it still requires changes to these protocols. These changes
> won't magically happen on their own and they won't happen in exact
> parallel.

Yes, it is possible to deply my approach as the interdomain QoSR
protocol. The protocol works intradomain without any magic. No
parallel effort necessary.

What does not happen magically is that intradomain QoSR protocols
interoperate each other.

It may or may not be possible to develop a intradomain protocol
which can interoperate with my QoSR protocol. But, such an approach
can be attempted only after the interdomaiin protocol is fully
defined.

> We can suggest that changes be made in intradomain protocols
> first where they can be studied easier.

As there is already a lot of study on QoSR, such as QoSPF, which
does not work inter-domain, we don't need more.

If there is anything to study about intradomain protocols, it's
the study on how intradomain protocol can interoperate with
the interdomain protocol(s).

That is, the interdomain protocol and its interface to the
intradomain protocol must be provided.

							Masataka Ohta