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: [PATCH] autoindex instead of index.html
Date Thu, 27 Sep 2001 22:11:50 GMT
From: "Greg Ames" <gregames@remulak.net>
Sent: Thursday, September 27, 2001 4:01 PM


> "William A. Rowe, Jr." wrote:
> 
> > HERE is the difference.  In 1.2-1.3, we have never called the normal
> > request cycle for subrequests, we've had hackish, occasionally broken and
> > unmaintainable variants for each subrequest flavor.  Several months ago
> > I joined those in common code.  Within the last month I rerouted all that
> > brokenness through ONE block of code (ap_internal_process_request) that
> > will draw out any discrepancies.
> 
> > I'd argue the right fix -could- be a translate_name RUN_VERY_FIRST hook
> > in the request.c module, that says, "Hmmm... a _file_ subreq, better skip
> > the translate name phase" and returns OK.
> 
> I like the sound of that.  Or why not have ap_process_request_internal
> do exactly the same test, and skip running the hook altogether?  That
> would save an extra computed branch in mainline code, and make life
> slightly easier for the CPU's branch prediction logic.

If we must skip the <Location > walk as well, this may be the better hack.

I'm entirely unhappy with the prospects of ongoing sub_req_lookup_file
bogusness.  But this seems like a bigger fish than Apache 2.0, and we want
to keep moving.

I'll commit a fix to write a note on the request, and skip the location_walk(s)
and translate_name as appropriate if that note is detected.  Look for the 
commit this evening, it will back out the original hack you've complained about.

Bill


Mime
View raw message