Because it works. Scripts/software that works does not experience bit rot. As long as it works, people will continue to use it.
Edit: OK, fine. It seems the downvoters are convinced that the code does in fact experience bit rot, and spoils with time. Probably like fish.
As a counterpoint, I think we can both agree that writing production software in one of the toy languages that pops up on HN frontpage every day is a bad idea. Writing in a language that is old enough that it has become niche has many of the same issues. It doesn't really matter whether the software is poorly understood because the language is new enough to be niche or old enough to be niche; the outcome is the same.
Perl is an old language, pre OOP. It's a language from the 90s, when everyone was hacking together code not knowing what the future would be like. There is a lack of organization. Perl is like C in syntax, features, and in speed. Python is like C++. Every failing you can find in Perl stems from it being a PP language first, from dirty code to kids not learning it in college.
Raku attempts to address every failing of Perl, yet be better than Python. If it succeeds, who knows, but I hope it gets at least a chance to make something of itself.
Would I write an API in perl? Nope.
Would I write something that massages the data that APIs spit out and processes it, drops into a queues, generates jobs, etc? Absolutely.
Mostly because I will write a few hundred lines of code for existing CPAN modules to solve the problem that I have, letting me to move on to doing something else. I like boring tools that just work. In 99.99% of the times the tools/whatever you are writing is not inline, does not need to have amazing performance. It just need to work and be braindead simple. Perl excels at that. I still daily use code that I wrote over 25 years ago and modified it a couple of times since that point ( integrating git ).
Edit: Can't reply below so I will reply here:
Re: I would not write an API in perl.
> I see you have not tried Mojolicious.
I have seen it, tried it, got excited about it and immediately came to my senses because I asked myself "Why on earth would I build or advise people to build an API in perl? It is slow. It is bad for concurrency. It probably has a pile of corner cases that I'm or some other people are going to discover. If I need to ingest something via web I will throw a simple nodejs app. Millions played with it actively in a "web" sphere, and hit most of the corner cases. There are thousands articles on stackoverflow and we have dozens of skeleton servers that already have all the needed foundations, including federated databases that support live password changes, traffic sharding, online promotion and switches, dynamic reloads, do-at-most-once queues, do-at-least-once queues, etc etc. Why would I want to reinvent all of this when the job of the API is to get the stuff and if the question cannot be answered via a couple of lookups push the request to a worker?"
Re: CPAN in 2019
> n 2019, I can only think you haven't looked around much recently.
Maybe. I have not seen it to be the case two months ago.
In 2010, that may have been true.
In 2019, I can only think you haven't looked around much recently.
Yes, Perl's threading support is terrible, and I do wish it was as effortless as in Raku. But you'd be surprised how often you actually only need an event loop and non-blocking implementations, and for the rest of the time there's Mojo::IOLoop->subprocess or IO::Async::Function to provide a thread-like model with efficient forking.
Why not? There are CPAN frameworks that make it a breeze.
The most efficent "modern" web development I have done has been with Perl.
Mojolicious, Catalyst, Dancer – take your pick.
Why don't you specify this single reason you allude to?
I think there are plenty of reasons, and they work to reinforce each other.
