Re: [Sidrops] Robert Wilton's No Objection on draft-ietf-sidrops-rpkimaxlen-15: (with COMMENT)

["Rob Wilton (rwilton)" <[e-mail]>]

Hi Ben,

Sorry, I should have replied sooner.

I think that your answer is fine, and you don't need to add any text.  Thanks for considering and for the explanation.

Regards,
Rob


> -----Original Message-----
> From: Ben Maddison <[e-mail]>
> Sent: 16 August 2022 17:30
> To: Rob Wilton (rwilton) <[e-mail]>
> Cc: The IESG <[e-mail]>; [e-mail];
> [e-mail]; [e-mail]; [e-mail]
> Subject: Re: Robert Wilton's No Objection on draft-ietf-sidrops-rpkimaxlen-
> 15: (with COMMENT)
> 
> Hi Robert,
> 
> Thanks for making the time to review.
> Some thoughts below...
> 
> On 08/15, Robert Wilton via Datatracker wrote:
> > I considered balloting this as a discuss (for a discussion), but this is
> > outside my area of knowledge expertise.
> >
> > Although the document indicates that the number of published ROAs
> should remain
> > the same, since each ROA can list multiple prefixes, was any consideration
> to
> > the potential increase in VRPs (if that is the right term) that this change
> > will cause and whether this may negatively affect routers that are
> consuming
> > the ROAs/VRPs?
> 
> I don't recall this coming up in the wider WG, but it was discussed
> initially between the authors.
> 
> The consensus from that conversation was that, whilst your observation is
> quite correct, the ability of the "global" VRP set to scale to DFZ-like
> size is necessary for ROV to be successful in the long term, and that
> this scaling requirement ought not to be met by sacrificing the
> effectiveness of its protections.
> 
> I am not aware of any implementors having indicated that the
> recommendations in the draft would pose a scale challenge if fully
> implemented.
> 
> If a scaling problem presents in future, my best guess is that it comes
> at the router-policy-evaluation layer (since bandwidth and cycles at the
> RP are ~cheap). All the "serious" BGP implementations that I am familiar
> with have solved the problem of efficiently applying large prefix-list
> policies - of which ROV-policy is really just a variant. All said: I am
> pretty sanguine about this!
> 
> Let me know if you think any of this needs to be in the draft itself? It
> feels more like a topic for a different document to me.
> 
> > Am I right in assuming that the number of valid ROAs that can be
> announced
> > should effectively be bound by the number of BGP prefixes advertised for
> an AS
> > and hence this shouldn't be a problem?
> 
> That's true, but since de-aggregation continues a-pace, it's not a very
> firm bound; it's not what I am betting on to allow us to scale.
> 
> > But other that the question above, I found this document to be very easy
> and
> > pleasant to read.
> 
> Thanks, much appreciated.
> 
> Cheers,
> 
> Ben