Re: [spring] CRH is back to the SPRING Use-Case - Re: Size of CR in CRH

[John Scudder <[e-mail]>]

Peter,

On May 29, 2020, at 10:36 AM, Peter Psenak <[e-mail]<mailto:[e-mail]>> wrote:

well, advertising the local CRH identifier for every node and adjacency
in the network from every other node is clearly a no-go from the IGP
perspective.

(Of course this objection only applies to the final (“distributed routing protocol”) bullet point.)

We’re recapitulating conversation that has been done on-list at least once. If I had time I’d find the reference and post, but as we know the conversation is hundreds of messages deep (at least!) and I don’t; sorry. Maybe someone else has a reference handy? If I recall correctly (and I may not) one exemplary solution to your objection is to make use of globally-unique (per domain, of course) identifiers, and yes, I do mean something semantically similar to a SR Node-SID. Other solutions are possible, depending on the use case — if the use case doesn’t require any-to-any connectivity within the domain, you potentially don’t need O(N^2) CRH identifiers to be present in the LS(P)DB even absent any kind of global uniqueness. Granted that any-to-any is the most general case.

Not to mention that the proposed encoding in
draft-bonica-lsr-crh-isis-extensions only allows one to advertise the
CRH identifier for a local prefix and adjacency, not for the remote ones.

That seems out of scope for discussion of CRH per se, unless you mean to say “this problem cannot be solved” instead of “this particular solution is deficient”. If it’s the latter, I think the LSR mailing list seems like a better place to take it up with the authors of draft-bonica-lsr-crh-isis-extensions.

—John