Re: [Sidrops] WGLC for draft-ietf-sidrops-ov-egress-00.txt - ENDS 11/25/2019 (November 25 2019)

[Chris Morrow <[e-mail]>]

On Mon, 02 Dec 2019 22:55:51 +0000,
Randy Bush <[e-mail]> wrote:
> 
> > I am somewhat puzzled, could you elaborate more what you mean by
> > "exogenous data".  Also I still don't see the "non-deteminable"
> > aspect?
> 
> prefix injected from static on A sent ibgp to B, and B has AS marking
> policy based on a community which may or not have been marked by A.

The general case here is likely that in a complex network (more than a
few routers), it's possible to have wonky things going on. Randy's one
example certainly is true, from a real network though I know of a case
where SOMETIMES prefixes get their origin-as set via initial creation
of the route (static route), and SOMETIMES the operator decided:

  "Oh, I should match on community-foo + attribute-bar and reset
   origin-as to BLOOP"

This leads to some hilarity when attempting to sort out RPKI/ROA data :(

  "Oh, that's clearly AS-BAR because we set that on the creation of
  the route!"

followed shortly by:

  "wait what? why is it AS-BLOOP??? WTF?"

In any case: "make the ROA you intend, all of them" is sound advice.
Verifying prior to shipping across your peering edge sure seems like
"being polite to the world" as well.