> I believe we discussed this last time but, with the new perspective, couldn’t 
> we just use the pair <IID, site-ID> to at least get per VLAN granularity in 
> some of these decisions.

Yes, that would work better since the site-ID is in the Map-Register (where you 
need the xTR distinction) and the IID is in the EID encoding of the EID-record 
where you need the VLAN association. So this is ideal. 

> We get the same benefits: no impact on the bis document, and some extra 
> granularity in the choice of DF or more detail when implementing split 
> horizon on the xTRs.

Right. Agree.

> And just to complete the story. When we use site-IDs this means that:
>  
>       • An EID registered with the same site-ID (and merge-bit) from 
> different xTRs is merged. If site-IDs differ this is considered a move.

Correct. But you have to deal with misconfiguration so 2 xTRs at the same site 
advertising a different site-ID doesn't look like a move. So the xTR-ID needs 
to be checked as well.

>       • For “discovery” purposes in multihomed groups: An L2 EID registered 
> from one xTR and a specific site-ID, needs to be notified to all xTRs that 
> are using that same site-ID

You have to keep a list of xTR-IDs for a merged registration over time. And the 
list is cleared when a new xTR-ID is discovered with a different site-ID. Right?

Dino

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

Reply via email to