Dino,
That is, for NERD it is all or nothing. But a typical ITR needs some
but not all mappings.
I didn't really think of NERD as an all or nothing solution. In fact,
one could easily envision use of NERD for some notion of "local" or
"popular", and use of ALT for The Rest. For instance, I could envision
a Chinese NERD that would be distributed in the China region, and ALT
used for the rest. Some understanding of traffic flows would be useful
in determining this sort of thing.
How about this:
(1) Have an ITR, when it boots up, ask its map-resolver for all
mappings it might have cached.
(2) It may not eliminate all Map-Requests but maybe a large number of
them.
(3) Does require the map-resolver to be a caching system.
Comments?
It sounds VERY good so long as cache expiration semantics are maintained
– that is, when a refresh timer goes off, no matter where the entry is,
it gets refreshed. It may also make sense that when a caching resolver
is dumping its cache, it reduce by a random epsilon the refresh time of
each entry so as to avoid synchronization problems.
But help me on two points:
What is the difference between the ITR asking a map-resolver for all
mappings and the map-resolver being co-resident?
Would it be possible for the map-resolver to itself pre-cache entries
that it has some reason to believe would be "interesting"?
Regards,
Eliot
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
