Yesterday, during a regional conference on free software, I attended a talk by Jesús González-Barahona on "Measuring Woody". This talk hasn't been published yet (AFAIK), so I'll have to point to previous versions of his project on counting the number of lines of code in the Debian distros. The previous version showed that Perl is 5th among Debian languages by number of lines, amounting to 2.41% of the lines. The same, more or less, goes for Debian 3.0, AFAIR. It's more than those other languages, but less than LISP (xemacs libraries rule) and shell (lots of installation scripts).
Jesús is looking for collaborators, but I was thinking about something else: applying his methods and scripts to CPAN, and measure results by namespaces, languages (well, only Perl and C here (and maybe XML) here), maybe even authors. The good thing is that lotsa more information is available on CPAN than in debian packages: authors, dates... anybody up to the task?
It would complement nicely the Perl at a glance site, at least