Because the situation around PHP has massively improved with PHP supporting real OO programming, huge performance gains and cutting cruft like register_globals and other insanities.

Um, I'm working on a PHP codebase at the moment. I hate the PHP standard library. It's awful and awkward, but my job is to make shit work, so I do, even if I hate every minute of it.

So much this. Plus, we have certified CentOS servers that are only allowed to run 5.4 for the next years to come. Not to mention that the syntax and stdlib are a nightmare.

Thank god, we convinced our bosses to rewrite our legacy systems in elixir, since ~10years of iterative php progress grinded to a halt with lasagna layers of spaghetti code of homegrown-everything. Unmaintainable now, but the feature request keep coming in, and will do for years to come.

Lucky you, I'm stuck with 5.3, because of old CentOS too.

That doesn't change the fact that a lot of people still hate on PHP like it's 2008.

"What do you mean it's not real object oriented programming? PHP5 has been out for TEN YEARS."

Probably because the haters never took the time to re-evaluate PHP.

I've written daemons, GUI programs (!) and even stuff like custom binary network protocols in PHP, in a fraction of the time it'd have needed in C/C++...

I wonder if it would have been better to have changed the name of PHP - maybe follow the fortran approach and tack on the year.

Can vs Should. While PHP is turing complete, there are much better toolkits for the jobs you described.

Sure, but PHP abstracts a load of stuff when you're doing cross-platform stuff - and what's more important, it allows you to keep one language (thus reusing codebase) when you're also doing stuff with web interfaces.

Neither Java, Ruby, Python nor anything else are as simple to setup and resource-friendly to run as a PHP-FCGI+nginx setup.

The guys behind MtGox were working on a PHP implementation of SSL, before it all went south there.

PHP5 didn't fix deranged fundamental stuff like ($x == $y) && ($a[$x] != $a[$y]).