Re: [Bgp-autoconf] Progress report of design team on IDR virtual meeting

["Dongjie (Jimmy)" <[e-mail]>]

Hi Jeff,

Thanks for sharing your opinion on this.  So you also think additional attributes about the peer should be communicated after the BGP session is established.

I agree that we need to define the minimal set of requirements of BGP autoconf/auto discovery, and you could check the list of common requirements in my previous mail.

Best regards,
Jie

From: Jeff Tantsura [mailto:[e-mail]]
Sent: Wednesday, March 25, 2020 6:53 AM
To: Randy Bush <[e-mail]>; Robert Raszuk <[e-mail]>
Cc: [e-mail]; Dongjie (Jimmy) <[e-mail]>
Subject: Re: [Bgp-autoconf] Progress report of design team on IDR virtual meeting

I’m with Robert here, autodiscovery meant to provide minimum necessary amount of data to establish peering relationship, the rest is done thought BGP.

Cheers,
Jeff
On Mar 24, 2020, 3:48 PM -0700, Robert Raszuk <[e-mail]<mailto:[e-mail]>>, wrote:

All,

I completely agree with Randy on this part:

" None of these example attributes are requirements, will be specified, given code points, etc. in the protocol design."

On the rest I would actually go even further and say that such things should be out of scope of pure and simple auto discovery.

Reason being that once you open a trunk - even if originally empty or containing only spare tire - folks will load it with junk beyond imagination making the interop nightmare, opportunity to keep old mantra alive fix one bug intro two etc ...

And as I said in the other note - some of this stuff being proposed should be transitive and IMHO belongs in BGP not in autodiscovery.

Thx


On Tue, Mar 24, 2020 at 11:38 PM Randy Bush <[e-mail]<mailto:[e-mail]>> wrote:
jimmy,

my apologies for my poor communication.  let me try with text

  Two prospective peers SHOULD be able to exchange arbitrary attributes
  to communicate things such as position or role in a topology, link
  speeds and/or loading, preferred time of day, favorite ice cream, etc.
  None of these example attributes are requirements, will be specified,
  given code points, etc. in the protocol design.  But there SHOULD be a
  protocol provision for communicating arbitrary attributes so that the
  operator may configure as needed for their environment.

hope that helps.

randy


--
Bgp-autoconf mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/bgp-autoconf
--
Bgp-autoconf mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/bgp-autoconf