Joe writes:
> There is no such thing as "on behalf of" in proxies.
> They cache things they get, and serve them to requests they receive.
>
> These definitions are dependent on who pays for it,
> who controls it, etc.
>
> And whether they're oblivious to them or not is
> implementation dependent.
>
> The use of forward/reverse still isn't justified.
>
> Joe
Unfortunately, the "on behalf of" section of the statement does impact
how a cache works. For example, the cost (in latency or dollars) to
reach an origin server is likely to be higher for a proxy that acts on
behalf of a campus or organization than for a proxy that acts on behalf
of a content-provider network. This means that the replacement algorithm
for the proxy acting on behalf of the content-provider network may be
different than the replacement algorithm for the proxy acting on behalf
of a campus.
I used the term "surrogate" in the IRCACHE BOF to try to capture the
idea that the device acted on behalf of a different network element
(origin server, in this case) and to make clear that there were
operational differences between that model and the model represented
by the current installed base of proxy servers (which acts on behalf
of a group of end-user browsers or other user agents). While I agree
that from a theoretical point of view they may be the same, they won't
be in implementation or operation; to me, that means they should be
distinguished.
I personally believe that the words "proxy" and "cache" have been
overloaded almost to the point of uselessness. The taxonomy draft has
to deal with that, and I am coming to the conclusion that the best way
may be to start off with the admission that these terms have been
fuzzied up by sloppy usage. Give the common definitions, fuzz and
all, then we can get on with more precise definitions which have some
rigor.
regards,
Ted Hardie
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:27 MST