Apart from trying to persuade them to use mod_perl ;-) have you had a look at CGI::SpeedyCGI? I've not used it, so I can't endorse it, but it seems to be close to what you're looking for and was described in an old YAPC 2001 paper. Another persistent Perl environment I'm aware of is PPerl. Another general approach to speeding up CGI is FastCGI.

References Added Later

• Common Gateway Interface (CGI) (wikipedia)
• FastCGI (wikipedia)

• Re^3: Can two separate responses be sent to the client's browser from Perl, such as via fork{}? by afoken (2023)

• CGI.pm on AIX 7.2 by anonymonk (2023)
• Review of CGI::Alternatives by anonymonk (2018)

• fastcgi without the module by morgon (2020)

Update: using this old node for CGI References nowadays

With the standard CGI model, where your web server forks off a process to run an external program, it's possible to do such a thing, but it seems really tricky and fragile. The web server doesn't necessarily know that it'll run a Perl program and the perl binary it launches almost definitely doesn't know that it would have subroutines already in memory somewhere. Even if it did, it would have to do a substantial amount of bookkeeping to bring that code in to the rest of the program.

Perhaps CGI::FastCGI would work more to your liking.

With the standard CGI model, where your web server forks off a process to run an external program, it's possible to do such a thing, but it seems really tricky and fragile.

It's not really that tricky or fragile so long as the external program knows it's going to run a perl program. That's how CGI::SpeedyCGI (a.k.a PersistentPerl) does it. It has a small binary front-end that looks for the persistent perl process, starts it if it doesn't exist already, and then hands off control to it. All you have to do is change your shebang line to point at that front-end. (And code for its persistent environment.)

-sauoq
"My two cents aren't worth a dime.";

As a fulltime Perl CGI programmer, I can tell you that you are going to have to think 100% backwards from mod_perl. For example, instead of trying to create the function once and only once, look towards generating them on demand.

For example, the package containing all of these functions can have a autoload that will create the function the first time it is requested.

It is hard to say for sure the best path without knowing how you are generating these functions, but also checkout the core modues SelfLoader and AutoLoader. They will delay the compilation of functions until they are first used.