IMO, Perl's downfall was mostly Perl 6, not the language.
Plenty people wrote plenty of Perl long ago. Yeah, the whole $ business is maybe a bit unintuitive, but it's the least of the problem really. It's easy to get past it.
IMO, the first part of Perl's downfall is that it didn't evolve fast enough. It was good enough that people tried to do big things in it and then it turned out it wasn't a good idea. Perl's OO for instance is kind of a neat hack that turns into a horrible mess with large projects. Large projects also increasingly need verification and safety because the debug costs rise higher and higher, and Perl is paradoxically safer at small scales. "use strict" works great in toy scripts and is nigh useless in OO-laden large projects where "strict" does nothing to check your $this->{foo}->{bar} hash trees. Yes, solutions sort of exist but they're all adhoc and you have to plan for them, and module writers don't use them, and...
That could have been survivable with the right improvements, but:
The second part is that Perl 6 was terribly planned and took bloody ages to get anywhere. People stopped writing Perl 5 expecting Perl 6 was around the corner, so why invest too much effort when it was clear 6 was going to be incompatible even early on? And it kept not coming out, so Python quickly ate its lunch.
Plenty people wrote plenty of Perl long ago. Yeah, the whole $ business is maybe a bit unintuitive, but it's the least of the problem really. It's easy to get past it.
IMO, the first part of Perl's downfall is that it didn't evolve fast enough. It was good enough that people tried to do big things in it and then it turned out it wasn't a good idea. Perl's OO for instance is kind of a neat hack that turns into a horrible mess with large projects. Large projects also increasingly need verification and safety because the debug costs rise higher and higher, and Perl is paradoxically safer at small scales. "use strict" works great in toy scripts and is nigh useless in OO-laden large projects where "strict" does nothing to check your $this->{foo}->{bar} hash trees. Yes, solutions sort of exist but they're all adhoc and you have to plan for them, and module writers don't use them, and...
That could have been survivable with the right improvements, but:
The second part is that Perl 6 was terribly planned and took bloody ages to get anywhere. People stopped writing Perl 5 expecting Perl 6 was around the corner, so why invest too much effort when it was clear 6 was going to be incompatible even early on? And it kept not coming out, so Python quickly ate its lunch.