[10457] in www-talk@info.cern.ch
RE: Mirror Negotiation
daemon@ATHENA.MIT.EDU (Brian Morin)
Mon Dec 9 19:11:20 1996
From: Brian Morin <bmorin@WPI.EDU>
To: "'www-talk@w3.org'" <www-talk@w3.org>
Cc: "'Koen Holtman'" <koen@win.tue.nl>
Date: Mon, 9 Dec 1996 18:54:17 -0500
Resent-From: www-talk@w3.org
Koen Holtman Wrote:
"It is not easy for a client to quickly determine which mirror which =
offers
the highest bandwidth. A HTTP protocol element for sending a list of
mirrors to the client only solves the easy part of the problem."
I actually had more in mind when I wrote the original post, however I =
thought I'd see how the basic idea would fly before getting specific.
If we assume that mirrors physicaly closer to the client are probably =
faster than ones further away or on another land mass, then including =
some kind of geographic location code with each mirror entry would help. =
The client could then experiment with the mirrors that appear most =
promising.
Koen Holtman Wrote:
"I believe that there can already be some mirror negotiation when =
resolving a
hostname to an IP address, but I don't know how effective that is."
My understanding of this is that it allows requests to be distributed =
among several servers on the same site. Perhaps this could be extended =
(I'm not very familiar with these protocols), however I think that you =
would have to mirror entire hosts and make the decision of what mirror =
to use at the server side. I already made the argument that determining =
what mirror to use at the server side is inferior to the client side.
Brian Morin
Student at WPI
Author of Fnord Web Server
