Christopher Morrow wrote:
> I don't know that you'd want an RS keeping information from you, as a
> downstream of that RS, would you? I'd rather see the things so I can
> decide what's best for me.

Chris, this is not how route servers work.

What you're suggesting will only work if the route server implements
add-path tx, allowing the client to receive all the candidate prefixes
for a particular route, not just the calculated best path.

In practice, what happens is that the route server calculates the best
path on a per-client basis and sends the best path to the client rather
than all the candidate paths.  In order to do this, the route server
needs to be configured so that the BGP decision process knows whether to
take the rpki-invalid or rpki-unknown prefixes into account when
calculating the best path for each client's Loc-RIB.  This configuration
needs to be done on the route server in advance, and on a per-client
basis, and can only be handled properly by config flags in the IXP
provisioning system.

> I think because this seems like a 'per network' or 'per ixp' concept,
> let's let the document not define the implementation, but just the
> capability.

This is the wrong way to solve this particular problem.  It's a local
IXP provisioning system flag, not an ietf issue.

Nick

_______________________________________________
GROW mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/grow

Reply via email to