Re: [RAM] Re: the separation of ID/RLOC

[Christian Vogt <christian.vogt at nomadiclab.com>]

Noel,

I was thinking of the "mapping function" as a mechanism that encompasses
both, the initial distribution of mapping information and subsequent
mapping updates (including cache invalidations).  I was looking at it at
a higher abstraction level.

When it comes to concrete, technical approaches, I fully agree with you
that a solution could realize the initial mapping set-up in a different
way than subsequent mapping updates.  First of all, this is because
requirements are different:  One may want to minimize signaling overhead
for the initial set-up, but minimize update delays, in particular in
case of provider fail-overs.  The solution spaces are different as well:
 E.g., current traffic patterns may allow you to prioritize mapping
updates for those remote edge networks with which a lot of traffic is
being exchanged.  Mapping updates could also be realized in-band with
regular traffic.  OTOH, for the initial set-up, neither prioritization
nor in-band signaling seems to be feasible to the same extent.

Where I don't agree with you is that the pull-push trade-off would only
apply to the initial mapping set-up.  I think the pull and push
paradigms are technical approaches that can be used for both, the
mapping set-up /and/ mapping updates.

Kind regards,
- Christian


_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram