Re: Initializing HTTP headers from HTML documents

• Mail folder: WWW Talk Jan 94-present
• Next message: neuss@igd.fhg.de: "Re: Passing info between pages"
• Previous message: John Franks: "CGI and typing files by suffix"
• In-reply-to: ellson@hotsand.att.com: "Re: Initializing HTTP headers from HTML documents"

From: Dave_Raggett <dsr@hplb.hpl.hp.com>
Message-id: <9401101008.AA15175@manuel.hpl.hp.com>
Subject: Re:  Initializing HTTP headers from HTML documents
To: ellson@hotsand.att.com
Date: Mon, 10 Jan 94 10:08:08 GMT
Cc: www-talk@www0.cern.ch
Mailer: Elm [revision: 66.36.1.1]
Content-Length: 1900

John Ellson writes:

> I agree with your suggestion that one way to maintain
> cache coherency is by a background polling activity prior to the
> document being re-requested from the cache.  This improves cache
> response at the expense of some background polling overhead.  I think
> this suggestion can be combined with Roy's proposal (see below).
...

Thanks for working thru the arguments nicely. I agree that EVAL would
only be useful for the primary server, and hence better done by making
using a script to generate the document instead.

>        If the document has an unexpired "Expires" statement in the
>        header then a cache server can immediately serve a document 
>        without checking back with the original server.

>        A document that has expired should not be served, instead a
>        new copy should be obtained from the original server.

>        A document without an "Expires" statement should be checked
>        against the source to see if the original document still
>        exists and is unchanged.  Cache implementers can do this
>        either as a periodic polling activity or at the time the
>        document is re-requested.  In either case the poll can be
>        done using Roy's suggested protocol.

This seems like a workable approach. All we need to do now is to
persuade people to modify their servers to implement Roy's check on
a "Last-Modified:" header in the client's request for HEAD and GET.
This can be done incrementally, as Roy points out:

> Note that implementing this protocol would have no effect whatsoever
> on existing servers and clients.  Old clients (and any without caches)
> would just continue making requests without Last-modified headers.
> Old servers (at least the NCSA httpd 1.0 that I use) will already accept
> a message of the above format and just ignore the Last-modified header.

Many thanks,

Dave Raggett