W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > May to August 1996

Re: Meaning of Last-Modified/I-M-S (was Re: Last gasp terminology issue)

From: John Franks <john@math.nwu.edu>
Date: Tue, 4 Jun 1996 22:08:06 -0500 (CDT)
To: Larry Masinter <masinter@parc.xerox.com>
Cc: mogul@pa.dec.com, fielding@liege.ICS.UCI.EDU, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Message-Id: <Pine.SUN.3.91.960604214158.1719A-100000@hopf.math.nwu.edu>
On Tue, 4 Jun 1996, Larry Masinter wrote:

> For web servers that are based on file systems, the "last-modified"
> time corresponds to the last time the file was modified, both for the
> case where there is a single file and when there are a fixed number of
> files and the appropriate file is chosen according to request headers.

Yes.

> 
> However, if you add a new file to the set, the 'last-modified' date of
> other members of the set would also change, so the CERN and Apache
> heuristic of using the maximum over all of the variants is a good
> conservative heuristic.
> 

I guess I would consider it a "bad conservative heuristic".  But I
think we agree that it is a heuristic and does not belong in the
protocol.  As you point out adding a new variant in effect changes the
last modified date of other variants (if it changes the set of headers
which will result in those variants being served).  Hence the current
draft which refers to the last modified date of the variant (not the
last modified date of a file) seems quite appropriate.  I am not up
with the latest revision of the definition of variant, but presumably
it includes not only the file served but the collections of headers
which can result in the choice of that file.

Even a good heuristic should not be required by the protocol, but
the maximum l-m-d of all the files of variants is a poor one, because
it is both inefficient for the server and "bandwidth unfriendly."
Any server which has resources with multiple variants must keep
a table of which variants correspond to which request headers.
Without having thought about it a great deal, it seems that a
better heuristic would be to define the l-m-d of a variant to
be the maximum of the l-m-d of its file and the l-m-d of this table.
Most likely others could come up with even better implementations.

John Franks 	Dept of Math. Northwestern University
		john@math.nwu.edu
Received on Tuesday, 4 June 1996 21:02:21 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:32:02 EDT