- From: Jeffrey Mogul <mogul@pa.dec.com>
- Date: Thu, 11 Jul 96 17:06:06 MDT
- To: Albert Lunde <Albert-Lunde@nwu.edu>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
But it does have one property I like: I'd say rather than just say
"choose the referal info with the best cost metric" I'd like
to say that clients should choose in a quasi-random way among
options with an equal cost metric. (What they do in the case
of unequal costs needs to be refined further to define what
this would mean.)
I agree; we should include an explicit recommendation to
randomly choose between equal-metric referrals. But we
need to think a bit harder about whether this choice should
be made per-request or "per session", however one defines
that. I suspect that if clients do it per-request, this
will play havoc with caches and with updates.
But I would also expect servers to offer referral-lists that
include referrals of quite different metrics. This would
be used to tell a client, for example, to use the "nearest"
site as long as it was up, but to fail over to a more distant
site if the preferred one became unreachable. It's not
inconceivable that a server could offer different referral-lists
to different clients, depending on each client's location in the
Internet topology.
-Jeff
Received on Thursday, 11 July 1996 17:16:59 UTC