Re: Concerning PR#2409 (customising ErrorDocument 500)

["Roy T. Fielding" <fi...@kiwi.ics.uci.edu>]

>PR#2409 complains that ErrorDocument 500 declarations in
>.htaccess files and <Directory> containers get ignored.
>I tracked this down to the way directory_walk() works: it
>parses the .htaccess files and then merges the config
>records.  Thus a parsing error (which means a 500) has
>only the server config available to field the problem.
>
>Does it makes sense to document this as a restriction, or
>should the .htaccess parsing include the merging operation?
>Personally, I prefer the latter -- any comments before I
>start working on a patch?

How can the .htaccess parsing include the merging operation
if the .htaccess file cannot be parsed?

My inclination would be to document it as a restriction rather
than adding cruft to workaround ErrorDocument 500 (which should
only be seen by an administrator testing a new config anyway).
OTOH, if partial merging can be done and is worthwhile for other
reasons, then you might as well try.

....Roy

Re: Concerning PR#2409 (customising ErrorDocument 500)

[Dean Gaudet <dg...@arctic.org>]

On Fri, 4 Sep 1998, Rodent of Unusual Size wrote:

> Roy T. Fielding wrote:
> > 
> > How can the .htaccess parsing include the merging operation
> > if the .htaccess file cannot be parsed?
> 
> The <Directory> and superiour .htaccess files that have been
> successfully parsed would be in the per-dir config chain.
> The failing one, of course, wouldn't be.  The merge would be
> the last thing the parse did before returning successfully.

Yeah I think all the successfully read .htaccess files should be merged in
before doing the error, but only if it's a small change.

Dean

Re: Concerning PR#2409 (customising ErrorDocument 500)

[Rodent of Unusual Size <co...@Apache.Org>]

Roy T. Fielding wrote:
> 
> How can the .htaccess parsing include the merging operation
> if the .htaccess file cannot be parsed?

The <Directory> and superiour .htaccess files that have been
successfully parsed would be in the per-dir config chain.
The failing one, of course, wouldn't be.  The merge would be
the last thing the parse did before returning successfully.

> My inclination would be to document it as a restriction rather
> than adding cruft to workaround ErrorDocument 500 (which should
> only be seen by an administrator testing a new config anyway).

Erm, no -- people do this sort of thing in [sub-]sub-directories.
Consider the 'fake' vhosting stuff that was done with mod_rewrite
a few months ago.. or ISPs that provide space for customers as
directories off the ISP's main URL.

> OTOH, if partial merging can be done and is worthwhile for other
> reasons, then you might as well try.

It may not be; I'll take a look.

#ken	P-)}

Ken Coar                    <http://Web.Golux.Com/coar/>
Apache Group member         <http://www.apache.org/>
"Apache Server for Dummies" <http://Web.Golux.Com/coar/ASFD/>