httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <>
Subject Re: seg fault in 2.0.23 :-(
Date Tue, 14 Aug 2001 02:31:50 GMT
From: "Roy T. Fielding" <>
Sent: Monday, August 13, 2001 5:48 PM
Subject: Re: seg fault in 2.0.23 :-(

> Nope, those look legit, although the coding style is poor.  This is going
> to require a complete rewrite.

Just some thoughts on mod_mime...

it groks many different things, but only 'succeeds' on content_type?

It's pretty bogus.  I'm thinking we need to ...

1. make the mime phase a run-all.  The modules can decide if they have anything
   to contribute.

2. pull the default language/content type/etc into a mime_core_default function
   that is HOOK_LAST.  Therefore ...

   mod_mime_fileinfo takes things out of extended file attribs where the filesystem
                     offers some support for charset/language/content type.

   mod_mime works over the filename extensions, further identifying the file.

   mod_mime_magic fills in the content type by scanning the file, if we still don't
                  have a clue.

   mod_mime_html  works over the html headers for good charset/language meta tags.

   core mime_default fills in the final, missing info based on per-dir defaults.

Some of these modules are hypothetical, but it does illustrate that the first might
pick up only a charset, mime might figure out html, but the language wouldn't come
in until mod_mime_html.

As long as one hook phase needs to fill in 5 variables (including the handler!) we
need to be run all.

And as I mentioned to Mr. Stoddard in a sideband headscratching, the table args to the
overlay_extensions_mapping function seem reversed.  I'm not touching this for a time,
however, since I have my hands full on some other plates.

But I'm back.  New Orleans was nice ('n hot), my brother's wedding was wonderful.
Sorry for my absense :)


View raw message