W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 1994

Re: HTTP Caching Model?

From: Daniel W. Connolly <connolly@hal.com>
Date: Wed, 14 Dec 1994 15:43:17 -0600
Message-Id: <9412142143.AA06326@ulua.hal.com>
To: "Roy T. Fielding" <fielding@avron.ICS.UCI.EDU>
Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
In message <9412132125.aa15482@paris.ics.uci.edu>, "Roy T. Fielding" writes:
>
>Since, in reality, only a small portion of URI space is capable of
>generating variants, I think a more general solution would be to have
>the server explicitly declare what variants are available (if any)
>as part of the response metainformation.

So back to my question of "where do we assign the fault?" you're
saying we should assign it to the origin server for not telling the
proxy that there were variants.

It seems like Mitra and Brian would agree.

Seems like a workable strategy, now that I think about it:

And I seem to recall that Tony Sanders gave an explanation* of the
URI: (aka Location:) header a while back that would solve this problem.

* http://www.research.digital.com/nsl/formtest/home.html

>  2) Patch the current model by adding a "Variants:" header like
>
>        Variants: text/html;qs=1;bs=9653;ua="Mozzilla",
>                  text/html;qs=0.9;bs=8753,
>                  text/plain;qs=0.2;bs=7989,
>                  text/plain;qs=0.3;bs=8989;lang="en/gb",
>                  application/postscript;qs=0.7;bs=90267
>
>     which would ONLY be sent when the given URI allows variants;


This is kinda humorous... the answer has been in the spec all along:

=======================
7.9 URI Header

The URI-header field contains a URI by which the object may be
found. It should not be confused with the token in the Request-Line
described in Section 5.4. As for a normal request, there is no
guarantee that the object can be retrieved using the URI
specified. The field is normally a part of a response having
Status-Code "301 Moved Permanently" or "302 Moved Temporarily".

URI-header      =       "URI" ":" 1#( URI [";" vary] )
vary    =        "vary" "=" <"> 1#vary-param <">
vary-param      =       "type" / "language" / "version" / "encoding"
        /       extension-vary
extension-vary  =       token
=======================


When a server returns an entity in response to a GET query on a URI,
it can give a URI header, which specifies the set of entities that
represent the accessed resource (at that time, or between the
Last-Modified and the Expires times).

We can describe the current practice by saying that when no URI:
header is given in response to GET xxx, the client should assume:

	URI: xxx

That is, the resource indicated by xxx has exactly one representation
(at that time).

If a URI yyy has representations that vary based on preferred content
type, language, and user agent, the server should return:

	URI: yyy; vary="type,language,user-agent"

The vary parameter tells a caching proxy exactly what headers it must
keep in its "cache-key."

This meshes happily with current practice, to a large extent.

The only changes requred are that

	* when a server returns a representation of a resource, if
	it has other representations of that resource, it _must_ give
	an appropriate URI: header.

	* caching proxies _must_ take URI: headers into account when
	looking for cache hits. (The quick and dirty implementation is
	to not cache any responses with URI: headers).


Hmmm... as I recall, TimBL or Tony S once explained that ideally, if
a server has a document in text, html, and postscript in english, french,
and german, and it's returning the french postscript, it should return:

	200 Document follows
	Date: Wed Dec 14 15:34:19 CST 1994
	Last-Modified: Wed Dec 14 10:34:19 CST 1994
	URI: http://this.host/path/to/multi;
			vary="language,type"
	URI: http://this.host/path/to/multi.ps;
			vary="language"
	URI: http://this.host/path/to/multi-french;
			vary="type"
	URI: http://this.host/path/to/multi-french.ps
	Content-Type: application/postscript

	%!PS-Adobe 2.0
	...

So the client user can make a link to:

	* the multiformat document
	* the postscript represenation (in any language)
	* the french representation (in any format)
	* the postscript, french version

And the caching proxy would create 4 cache entries.

This is the ideal case. At a minimum, the server must return:

	200 Document follows
	URI: http://this.host/path/to/multi;
			vary="language,type"
	Content-Type: application/postscript

	%!PS-Adobe 2.0
	...

And the caching proxy could just not cache that response.

Dan
Received on Wednesday, 14 December 1994 13:55:10 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:09 EDT