[Idr] Re: New Version Notification for draft-kriswamy-idr-route-type-capability-00.txt

[Jeffrey Haas <[e-mail]>]

> On Nov 8, 2024, at 6:01 AM, Robert Raszuk <[e-mail]> wrote:
> 
> Jeff, 
> 
> > Ah, yes. Avoids sending routes to peers in the distribution graph that don't want them.
> 
> You are still confusing filtering from RR to PE (side product which even RT ORF could solve)

RT ORF was killed because RFC 4684 supplanted the use case and did so much better.

> from PE to RR or between ASNs where RT information on senders pertaining to receivers would not be there in the first place. The goal of constructing the graph and primary trigger for RTC was to automatically build disjoined RR cluster RT membership wise as well as prevent unneeded VPN routes to be propagated between ASNs doing VPNv4 peerings. 
> 
> This is precisely what RTC RFC solves. 

It's a flooding graph protocol. The graph is used to control where the routes are sent amongst the participants. 

You haven't said anything that contradicts that.


> 
> > So, if you find that required behavior generally problematic, you have an issue with that RFC with your name on it.
> 
> You are simply wrong here on the primary goal of RFC4684 - please kindly refrain from continuing misstating it. 

Sorry, who is trying to silence the mic now?

> 
> You can stay with thinking that sender side filtering is always much better then inbound drops. Your choice. 
> 
> And sure we have ORF, RTC let's now add capabilities at OPEN to carry NLRI filtering. Great for BGP - no doubt :(. 

Certainly. It's come up in discussion about all sorts of inappropriate route distribution over the last several years:

- Flowspec for determining filter support.  A problem that will need to be addressed in FSv2.
- Attribute escape issues for features that should be working in constrained domains.  Most recent example of the problem, DPATH.
- And now, incremental deployment headaches with typed NLRI.

I think we've otherwise exhausted our own technical discussion.

You're encouraged to discuss your further concerns with the other authors on the draft.  You may find at least one of them on one of your internal distribution lists.

-- Jeff