actually, I was thinking of mentioning this too.

In the end you only need a heuristic to recognize perl-code (easy) or data (consecutive short lines), this heuristic can be trained with the 100_000 older posts with code tags.

> But half the requests in SoPW are too badly written to be recognized as Perl even by the perl interpreter.

It's possible to run Perl::Tidy in server mode, i.e. without start-up overhead its pretty fast.

It's also possible to add an extra display-format where the reader decides between raw and formatted code.

But that's all fantasy, there are only two gods left who keep this running.

Pmdevs were bugging them already for years to patch a serious JS backdoor in the markup format.