Re: [Sidrops] draft-ietf-sidrops-rov-no-rr

[Job Snijders <[e-mail]>]

Hi all,

I'd like to start off by offering my compliments to the creators of this
useful concept. I also appreciate the generosity of making the
specification available through the IETF process rather than the US
patent office! :-)

As this Internet-Draft doesn't have an "interopability aspect", rather,
it describes an implementation aspect not visible to external parties,
building an interop matrix doesn't seem applicable.

On Wed, Apr 06, 2022 at 10:28:42AM -0700, Randy Bush wrote:
> > Was there some code already that implements the no -route-refresh
> > bits you want?
> 
> ios/xr.  jakob did it as soon as he saw the draft.

Small request for clarification: you mean IOS XR and vanilla IOS? Or IOS
XE?

> > (do we know the implications of doing this at scale?)
> 
> we have seen the implications of not doing this at scale.  ask mark
> tinka, who was depeered by at&t.

Agreed on this being a problem at scale. At a previous employer we also
ended up inadvertently imposing significant strain on neighboring BGP
networks by virtue of sending out ROUTE REFRESH messages more often than
strictly needed. We received various complaints from EBGP peers because
of this. The situation was somewhat ameliorated by running the
validation process three times an hour (rather than every minute, or
even every 10 minutes), but utimately the solution was to instantiate a
fully populated Adj-RIB-In for each EBGP peer. As the I-D mentions,
enabling 'soft-reconfiguration inbound always' (or its equivalent) can
pose a challenge in resource constrained environments.

Other than some editorial polishing, I think this draft is good to
proceed in the publication pipeline.

Kind regards,

Job