[rr-fs] RE: [Irtf-rr] New subgroup being formed

[Dmitri Krioukov]

brad,

please see below.

thanks,
--
dima.
http://www.caida.org/~dima/

> -----Original Message-----
> From: Brad Passwaters [mailto:bjp at theinternetisfull.net]
> Sent: Friday, February 27, 2004 9:18 AM
> To: avri at psg.com
> Cc: Dmitri Krioukov
> Subject: Re: [Irtf-rr] New subgroup being formed
>
>
> On Fri, Feb 27, 2004 at 10:14:50AM +0900, avri at psg.com wrote:
> > Those who wish to comment are welcome to do so either on
> > the list or privately to Dima and/or me, as they wish.
>
> Things that came to mind reading the charter:
>
> 1) Goal 3 - In addition to documenting the selected algorithms perhaps
>    a draft/section that pulls together why the other algorithim that were
>    considered were not selected.

ok.

> 2) Goal 2 - construction of realistic ("explanatory") Internet topology
>    evolution models (topology generators) - is this looking at
> things at the
>    AS level, for purposes of generalizing the building blocks that make
>    up the Internet or something else?

yes, it's at the as-level. the reason the topo research
is in the charter is that the performance of a routing
algorithm is a function of the network topology. so,
if we do not know the internet topology well enough,
then we might be wrong saying that a pre-selected
routing algorithm will perform well in the internet.

> 3) Will there be any consideration to the policy aspects of routing
>    currently in use? The policy needs certainly seem to affect message
>    size,routing table absolute size in terms of memory usage and
>    computational power.

as far as i know, policy aspects will be handled in a
different subgroup. at this point, it's an issue separable
from the fundamental scalability aspects of routing
(for fundamental aspect of policy routing, please see
tim griffin's works). at later stages, these two
efforts should be coordinated, of course...

> 4) Is this strictly for EGP type protocols or will you be looking at IGP's
>    as well?  If just EGP types will you be looking at the uses of the EGP
>    as an IGP or its interaction with an IGP as a criteria.

this is not a requirement, but yes, i think it would beneficial
(in some sense :) if the end result would look like egp = igp + policies.

> Not sure all this is in line with what you were thinking but its what
> I was wonder while reading the new charter.
>
> Brad