W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 1995

Re: Cacheable extension methods (was: an idempotent idea)

From: Larry Masinter <masinter@parc.xerox.com>
Date: Sun, 10 Sep 1995 22:54:31 PDT
To: sjk@amazon.com
Cc: fielding@beach.w3.org, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Message-Id: <95Sep10.225436pdt.2763@golden.parc.xerox.com>
Roy said:

	A proxy cannot forward a method it doesn't understand.

And Shel waded in, but I thought I would throw in a few punches, too.

Proxy's most definitely can forward any method that they get. For
example, a simple proxy might react to

<method> <fullurl> <httpversion>
<headerstuff> 
blah blah
 ...

and parse <fullurl>. If <fullurl> doesn't start with "http://", it
must of course understand <method> so it can translate the protocol.
Otherwise, it could just open a TCP connection to <fullurl>'s host and
port, and send 

<method> <restofurl> <version>
<headerstuff>
blah blah
 ...


down the TCP IP connection, and return whatever it gets on the way
back, and close the connection to the other side whenever one side
closes the connection.

This works for things where the connection stays open, and all kinds
of new methods.

Of course, proxies have to understand the semantics of <method> and
<httpversion> in order to do CACHING. Yes, you need to know what
someone really intended before you start saving away the response and
feeding it back to clients later.
Received on Sunday, 10 September 1995 22:56:41 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:31 EDT