From: "Joshua Slive" <[EMAIL PROTECTED]>
Sent: Monday, August 20, 2001 11:04 AM
> On Mon, 20 Aug 2001, William A. Rowe, Jr. wrote:
>
> > But because I'm building a 'composite negotation typemap/content' method, where
> > the new sytnax will be;
> >
> > Content-mechanism: values
> > Body:end-tag
> > <p>This is actual content</p>
> > end-tag
> >
> > I'm suggesting we leave the contribution where it is, and build these composite-ed
> > type-map/content files right into httpd-2.0/docs/error/.
> >
> > Would this be acceptable to everyone?
>
> I'm slightly skeptical about the usefulness of this new typemap thingy.
> Is the point of this just to save a bunch of "stat" calls?
> I guess I could be convinced, but it seems less clean to me.
Derriding "Just saving a bunch of stat calls" is a way to get lynched by new-httpd :)
Seriously index.html has 28 language/charset variants in apache-1.3, 35 in httpd-2.0.
Let's call it 40 by year-end.
There are 30 ways for a request to die with a 4xx/5xx response.
Let's say 1200 files :(
The most painful thing to mod_negotation (or mod_autoindex, for that matter) are
directories
that are _too_large_. The most painful thing to the stat call is, the same.
So instead of 1200 files, we have 30 files. Each one serves a specific error, so that
file
is accessed directly. No multiviews, one file read, and the error is served _right_
_out_
_of_ _that_ _file_!
It's essentially manditory. Anyone who has p/w protected (or removed) a huge chunk of
content,
only to watch 50 spiders come in and fail on their refresh, would be very glad that
these
not found/auth required requests didn't consume much cpu/memory.
Bill
