httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@covalent.net>
Subject Re: Regarding 1.3.19...
Date Fri, 23 Feb 2001 19:48:55 GMT
From: "Dietz, Phil E." <PEDietz@West.com>
Sent: Friday, February 23, 2001 11:57 AM


> What I planned on submitting was adding a per-directory directive called
> 'DefaultHandler_UseInodeInEtag'.
> This fixes the common-man's problem.
>
> It's probably not the solution for the "big picture."
>
> Reading the w3.org doc on ETags, it makes you feel that there should be a
> unique make_etag() function for every Apache handler (instead of just one).
>
> Since autoindex has it's own handler, shouldn't it emit it's own ETag ??

No ... I believe we could end up with an identical result from a number of
different mechansims.  The spec is rather specific, that ETag only applies
to that particular URI.  It cautions proxy authors against trusting that the
same etag for two different URI's represent the same resource.

The etag on autoindex comes from the directory's own inode, and that is
built into the mod_autoindex handler (based on the IndexOptions directive.}

I'm guessing that an Options DisableINodeETag would possibly be the best comprimize.
It could be set against a specific directory (a mount point of an inode-less
or untrustworthy inode'ed network volume.)  It could be set at root for your
scenario.  And it would be combined with the IndexOptions tag, such that the
IndexOptions could be set globally to report inode in the ETag for platforms
with trustworthy directory inodes, and the Options NoETag would disable the
behavior.

In other words: Options DisableINodeETag eliminates inodes from the default
handler's behavior.  Options DisableINodeETag overrides the mod_autoindex
IndexOptions' etag.

Once we have concensus, I'll do the mod_autoindex side of this patch.

Bill








Mime
View raw message