I can use some hard figures: HP-UX 10.20, considered dead, abandoned by HP years ago. I still make perl available in software depots for this OS on my website for users of this OS. Let's call these "innocent users". They probably have to maintain this old OS for whatevere obnoxious reason, but apparently they do. Only this year, perl has been downloaded for HP-UX 10.20 602 times. Of those, there were 5 for 5.12, 25 for 5.10.1, 19 for 5.10.0 and 34 for 5.8.9, so these people are still interested in recent versions of our favorite scripting language.

There is no C99 compliant version of HP C-ANSI-C available for HP-UX 10.20, and there never will be. You need HP C-ANSI-C to build GNU gcc on HP-UX (or a pretty recent version of gcc that someone else built for you). That is why I also make builds for gcc available. Building gcc-4 is absolutely impossible on HP-UX 10.20.

If these people want to install a CPAN module that uses XS code, they need to have a C-compiler installed. No C99 available for HP-UX 10.20, so end-of-story.

Now back to your claim: of those hundreds (or thousands) of Perl module authors, how many use XS code? And of those (which is in my guess way below 30%) how many use (or want to use) specific C99 features?</P.

In that perspective, isn't it reasonable to ask politely to consider using C89, and not beyond? How many limitations are you aware of that would make your XS coding less fun? (besides the silly // comments that should never have been added to ANSI-C: C is neither java nor C++)