Re: scope of solution..

[Brian E Carpenter <brian@hursley.ibm.com>]

RJ Atkinson wrote:
> 
> At 10:29 12/04/01, Brian E Carpenter wrote:
> >RJ Atkinson wrote:
> >>
> >> At 12:28 11/04/01, Brian E Carpenter wrote:
> >>
> >> >Possibly, but that would also require the site's routers to have
> >> >knowledge about the state of IDR, in order to give useful hints
> >> >to the hosts. This is not trivial.
> >>
> >>         However, it is already the case for multi-homed sites
> >> that the site's edge routers have to meaningfully participate
> >> in BGP4, so that requirement would not be worse than the
> >> currently deployed multihoming approach...

I don't disagree with that, and it adheres to the "first, do no
harm" principle which I've been advocating.

> >
> >IDR is not necessarily BGP4 on the timescale we are discussing,
> >and I'm not sure that BGP4 conveys adequate information anyway.
> 
>         The deployed multihomed sites in the Internet
> represent an existence proof that the information carried
> in BGP4 is adequate for multihoming purposes (regardless
> of whether one likes BGP4 or not).

Minimally adequate perhaps, but surely there is a lot more
information in the topology than is actually used today.

> 
>         One might imagine replacing BGP4 eventually,
> but that would be well outside the scope of this particular WG.
> In short, one can construct hypothetical scenarios unrelated
> to the real Internet, but I don't see how that is at all helpful
> here on this topic in this IETF WG.

I think the relevance is this: if the routing system is to give
hints to the hosts, these hints should *not* be tied to BGP4 specifics
in any way (i.e. they should be more generic than prefix length or
AS path length).

Regards
   Brian