Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@httpd.apache.org
Date: Mon, 1 Oct 2001 09:22:48 -0700 (PDT)
From: Dirk-Willem van Gulik <dirkx@covalent.net>
To: <dev@httpd.apache.org>
Subject: Re: .asis handler isn't driven
In-Reply-To: <021501c14a8c$b4881aa0$94c0b0d0@roweclan.net>
Message-ID: 
 <Pine.OSX.4.33.0110010911070.535-100000@titatovenaar.sfo.covalent.net>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


On Mon, 1 Oct 2001, William A. Rowe, Jr. wrote:

> We still have no mechanism to 'quantify' the quality of one handler or
> filter over another.  Funny that
>
> index.html  4000 bytes
> index.cgi    500 bytes
>
> will serve index.cgi, based on size, when the cpu impact of index.html is
> _significantly_ less expensive ;)

Well length is last -if all other things equal-; sourcce_quality is first.
since you can associate quality with any mimetipe/extensions. And this
will help you out for the common case. (As does the hardcoded 'set quality
to 2 (or some other high value) on CGI_MAGICTYPE) when we have args,
a post or some other indication of CGI).

Having said this this may sound like a solved problem - this is however
not a generic solution. As this variance space is not a normal 'ordered'
space where 1>2 and 2>3 and you got nicely spaced values. In fact this
space behaves more like a logarithmic one - where distance denses - and
hence we have all those fiddle factiors; the 0.001f's and so on. Perhaps a
rewrite is in order -where we use sequential ordering per mime dimension
(i.e. lang,charset,encoding,lenght) with a *linear* cost function.


Dw