Re: [Idr] I-D Action: draft-ietf-idr-large-community-01.txt

[heasley <[e-mail]>]

Thu, Oct 13, 2016 at 12:46:28AM +0200, Robert Raszuk:
> >
> > Thu, Oct 13, 2016 at 12:08:44AM +0200, Robert Raszuk:
> > > I believe Jeff was talking about intelligent cleanup with sufficient
> > > information embedded into community itself such that forwarding BGP
> > speaker
> > > can decide what to pass and what to drop
> > > ... not an unconditional blind drop.
> >
> > I (we, I believe) do not want vendors to do anything automatic about
> > dropping/altering communities.  But, provide us all the tools to manipulate
> > them.
> >
> 
> ​It's like asking vendors .. Dear vendor pls give me tools to unlock
> rounded by design metal nut :) ​

I do not understand that analogy.  Automatic dropping is like giving me
the toolbox full of tools but no key to unlock it. /)  btw, tools exist
to remove rounded nuts.

> > > Some say that standardizing, deployment, education of large communities
> > > will take years. And in fact they are very correct. It will take that
> > much
> > > indeed.
> >
> > I think you are speculating to push your drafts; you are certainly
> > persistent.
> 
> No .. not at all. I currently gave up on pushing wide or even common header
> idea if you have not already noticed since large was accepted as WG doc. ​

yes, you've stopped trying to block -large- for -wide-, but no operator
suggested that you stop pushing -wide- for the sake of -wide-.  I think
-wide- is a good idea.

> > I do not expect education to take much time at all.
> 
> ​Well I value your expectations however the personal experience educating
> network community in Nanog, RIPE, APRICOT and many *NOGs proves me
> otherwise. I think Job was right on the spot in his recent mail on the
> estimate of the effort.

they know rfc1997; this is rfc1997+.  unless their vendor does something
silly with RPL features, where is the difficulty?

> > And, I expect that we will see rapid adoption (deployment) in
> > some specific places; particularly exchanges with RSes.
> 
> ​If you need large community for IXes then it is fine. To me real use of
> BGP as protocol is a bit more then that.

I'm suggesting that there is a need there, a clear one with great benefit
to them (ie: incentive), and RSes have more flexibility to alter their
code and upgrade than others.