Re: [Idr] draft-ymbk-sidrops-rov-no-rr

[Jeffrey Haas <[e-mail]>]

On Mon, Nov 15, 2021 at 09:16:27PM +0100, Job Snijders wrote:
> On Mon, Nov 15, 2021 at 02:58:50PM -0500, Jeffrey Haas wrote:
> > Randy's other commentary seems to be emphasizing the main part of the
> > pain is effectively self-inflicted: If you quickly and consistently
> > distribute route validation updates in a consuming network,
> > implementations that are using route refresh to deal with those
> > updates by doing what they're told - issuing refreshes.
> 
> I concur it partially self-inflicted, but keep in mind that any EBGP
> peers receiving those ROUTE REFRESH messages as a result of this
> phenomenon didn't ask for refreshes. Operators are learning about this
> infliction 'the hard way'.

That's understood.  This isn't any different than any other over-aggressive
provisioning tool's impact.  In this case, it just happens to be coming in
from a protocol.

As this thread points out, the industry seems to be moving to more and more
aggressive distribution of the RPKI state.  While we often talk about
"boiling the ocean" in terms of trying to solve too large of a problem,
we're highlighting that we may be trending the other way.  Given an
aggressive enough ROA distribution infrastructure, a single update may cause
the Internet at large to burn CPU to deal with the route re-evaluation
impact.  

A pebble tossed into the ocean and causing it to boil is likely not the
intended impact.

-- Jeff (the IGP people are likely laughing at us now...)