I feel that Mason is too much tied into the Mason approach. I didn't
think it was as extensible or easy to override as some other options.

I'm running my own MVC framework, but if Catalyst were stable when I
started work, I would have chosen it.

Personally, I only use TAL for view components. Petal compiles it
into fast enough Perl, and Template::TAL works if you don't want
caching. I've run into a few bottlenecks in code, and have been able
to keep the same templates and just proxy the URL to a python daemon
or a quick PHP page ( you're thinking php- he must be crazy. but
yes, php. when you're dealing with data on a certain URL thats
unpredictable and can bump up your mod_perl server by 20mb for a
single request to process, you can proxy only that url to a single
fcgi php handler on lighttpd that takes up 3mb of memory when
dormant and spikes to 20 then releases ASAP unlike mod_perl when
needed ) Little stupid stuff like that makes TAL my choice for views
in any language.

If you're looking at putting raw-perl into a page... i think embperl
would probably perform better than Mason, as it looks to have less
magic going on than mason (ie, embperl is more of perl in html, and
mason is more mason code in html )

You can also look at danga's Perlbal. its supposedly fast as hell,
but like all their stuff: undocumented

-- i just saw the next message you wrote. I'd probably strongly
suggest having people write for you in TAL. serve it via modperl or
serve if on ( lighttpd + fcgi + php + eaccelerator ( sounds like a
lot of overhead, but its a fraction of the apache resources and
outperforms apache by 20%))