You'll have to pry it out of my cold, dead fingers!

One example why: I have a client whose on-line ordering system is pure CGI (well, there's a little js, purely for rendering on the "order" page itself. The system -- login, verification, presentation of available stock in an order-sheet format; calculations including volume discounts; billing; and order confirmation to client & customer -- is probably hit about 10 - 35 times per business day. At an average thousand dollars a pop, my client is quite happy (particularly since a significant part of his customer base is computer-phobic)... and -- as we all know, a happy client makes for a happy dev.

It eats almost no cycles, costs the client no significant o/head and "just works," as it has since being debugged nearly a decade ago.

Granted, that's not a major argument for retaining CGI.pm in core... and, granted (again), were I just starting today to scope out a system for this particular case, I might go another route. Hypothetical "best practice" (because best practice in Case A may not be so in Case B) considerations don's cut it with me; even less so does any sort of "code purity" (or is it philosophic correctness).

P.S. It would be a major misunderstanding to read this as a slap of any sort of chromatic or any other advocate for improving Perl. It's not and I do admire such well-considered advocacy.