Our company does web hosting, and every single time we get an enquiry involving PHP it comes with a caveat along the lines of "but I need version 4.x.y with version z of the q module and safe mode turned off". The best one yet was someone who wanted us to promise never to upgrade PHP because his XSLT module needs pre-version 4.4 - a bit of googling revealed that there's a whole section of the PHP community hunting for servers that will never upgrade for that reason. We never ever get this with perl, because backwards compatibility works over decades.

Perl is stable -- Perl 5 has been out what... 10 years now? CPAN is also an amazing module repository and makes installing modules a breeze. Perl is also more powerful than php, and the syntax allows for certain things to be done more succinctly.

The best thing about PHP is that it's easy. The syntax is simpler than Perl, so it is easier to pick up. It also has a ton of built-in functionality so you don't have to go looking for modules as often, but that mass of bundled functionality combined with the fact that PHP doesn't have namespaces makes for a mess, compounded by the fact that the included functions are often named inconsistently.

Having done hundreds of thousands of lines of code in both languages, I favor perl; but don't hate PHP either. Both have tons of documentation (PHP's being more newbie friendly, but Perl's being more extensive). Both are very fast when coded properly.

I think this is what people mean when they call the management and design of the PHP language 'amateurish'. Blatant lack of good engineering practice.

Totally agree 100%. Another example: did you ever use nl2br() [php.net] to convert newlines into <br> elements? It's an extremely common thing to do. In a minor patch release, they changed the function to generate XHTML instead of HTML. In one stroke, everybody who thought they were generating valid HTML had errors in their code. This might not sound that bad, until you realise that nl2br() can appear a lot in large projects, there's no way to get the old behaviour of nl2br() back, and if you have a decent QA process in place, you'd be being notified of the errors across all the websites you maintain. You end up having to go back and change all your code to use generic string replacement functions.

Now, maybe you might say that it's a sensible thing to change (I disagree, there should be different functions for HTML and XHTML), but at the very least, they should have put a change in semantics in a major version update, not sneaked it in between 4.0.4 and 4.0.5.

It's not really the design of the language that's the real problem (although it's not pretty), it's the cavalier attitude from people who don't seem to take a professional attitude to their work that really grates.

Amen to the cavalier attitude. You know about PHP's Javascript-esque === operator? (that's the one with three equals signs). That got designed on the spot in an IRC session with Zeev and some other devs. Because I actually had to explain to these folks what the concept of "object identity" was, i.e. what lisp does with 'eq', python does with 'is', and Javascript does with ===. Yes, because PHP's is different. Not only does === fail to do object identity testing, it's simply '==', does all the "deep comparison" of ==, but also bothers to compare the type.

In other words, I was unsuccessful in explaining this rather basic concept. They got it blisteringly wrong, and hacked this wrongness into the language for all time. I attempted to explain (much more patiently than here) that no, this is not what === is supposed to do, but I wasn't heard. Not by Zeev, not by anyone else on channel. No one got it at all.

I passionately hated PHP for a long time after that, but it's just not relevant enough to my work anymore to hate. I have choices, I don't have outside clients demanding I use PHP anymore, and my choice of languages is respected where I work. I've chosen python for some projects, perl for another, C++ for another (the C++ one was of course not web). I could probably write my next project in Haskell and no one would bat an eye (though I will be stuck with maintaining it for all eternity). I'm even eying Prado -- a PHP library -- for an upcoming project, though I've still no desire whatsoever to write actual business logic in PHP, so it'll have to be solely at the view end of things.