httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Stein <gst...@lyra.org>
Subject Re: [patch] mod_access/3821
Date Sun, 21 Mar 1999 00:06:55 GMT
Cliff Skolnick wrote:
>...
> As for "still allow", before this patch you were getting random behavior.

right.

> Now the right thing would be to probably add another method that you could
> use that would give you the behavior you need, and increase the size of the
> array.  This is the long term fix for the bug, any my patch will not
> interfere with any valid method.

I think the long-term fix will be something quite different. Third
parties cannot introduce new methods and have the various Apache
facilities apply to them. You must be an Apache developer to add a new
M_FOO method. I ran into this problem with mod_dav, but (thankfully) Roy
added the DAV methods into Apache. However, there are new, upcoming
methods (not yet standardized) that need to be recognized. Basically,
I'm back where I started :-) Even though I can now go and tweak httpd.h,
that can't/shouldn't happen until a new RFC happens, and it still
doesn't help third parties.

The method recognition must be pluggable in some way, and the
cmd->limited bit field needs to go away. For the former, I typically
recognize M_INVALID and follow that with a strcmp() on the string
itself. For the latter, it should probably be replaced by a table
mapping method to a flag (or possibly its presence in the table meaning
that a limit is being applied to that method).

There are probably other areas to tweak for this, but in general the
hard-coded methods presents particular problems for module implementors.

Cheers,
-g

--
Greg Stein, http://www.lyra.org/

Mime
View raw message