Re: Faster HTTP Was:Re: The Superhighway Steamroller

Larry Masinter (masinter@parc.xerox.com)
Sun, 10 Jul 1994 22:43:02 +0200

If the URNs you send in the protocol are allowed to refer to mutable
data, the server can't easily maintain a valid cache. Isn't it better
to have a kind of URN that references data that cannot change out from
under it (e.g., a MD5 hash signature of the headers?)

In general, using hash signatures for data allows lower bandwidth and
processing time, e.g., you could send some combination of
accept headers
credentials

and other elements, hashed with the key of the client; the server, if
it hasn't either recently authenticated the client or parsed the
accept headers could return back to the client a request for more
details, which the client would resend.

The server could then keep this and hash it and use it in subsequent
retrievals.

This would allow a relatively low bandwidth transaction for most
subsequent events, and yet avoid requiring the maintenance of state
for arbitrary amounts of time.