At 03:04 PM 11/19/2000 -0800, Randal L. Schwartz wrote:
> >>>>> "Didier" == Didier Godefroy <[EMAIL PROTECTED]> writes:
>
>Didier> on 11/19/00 4:55 PM, Ime Smits at [EMAIL PROTECTED] uttered the following:
> >> | How do we make sure regular CGI scripts are using mod_perl???
> >> | Is there a way to find out?
> >>
> >> Check $ENV{MOD_PERL}, it should read something like "mod_perl/1.24".
>
>Didier> Can I assume that whenever mod_perl is enabled that all existing perl
>Didier> scripts use it without any modification of any kind?
>
>Not at all.  Apache::Registry is a good stop-gap, but you have to play
>by a number of useful and documented rules.
>
>When you're *really* good with mod_perl, you abandon Apache::Registry
>and move up to writing handlers or using embedded templating systems
>like EmbPerl, Mason, or Template Toolkit.

I do not consider Apache::Registry a stop gap. Maybe something like 
Apache::PerlRun is. The "rules" that you have to play by are also quite 
similar to the rules you would have to play by when writing a handler anyway.

Apache::Registry programs are basically compatible with Velocigen, PerlEx, 
and SpeedyCGI acceleration. I think that's a huge commercial and open 
source advantage for anyone writing web programs for sale or open community 
use.

Most people who write applications that require the use of mod_perl as 
opposed to using mod_perl as an option are shooting themselves in the foot 
as much as people who write applications that can only run in CGI 
mode.  There are occasional exceptions of course (eg AxKit makes reasonable 
use of Apache as a low-level system).

Most normal applications do not require every bit of speed eeked out of 
them that may be possible with an Apache handler. In the mod_perl guide, a 
reasonably fast machine (which most production mod_perl sites run on) shows 
very little time difference between a real-world app (what I would 
categorize the heavy Apache::Registry vs heavy hander) running in 
Apache::Registry versus being written as a handler.

The use of Apache::args and Apache::Request::param seem to provide a better 
bet for the code optimization money than simply moving out of 
Apache::Registry to a handler().







---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to