The prehistoric luddites and trolls on the list would shout down any truly language-improving proposals in the same way they always do, regardless of source.
The only way to get rid of those unwanted elements would be a fork, but that would be unproductive. The remaining thing keeping PHP so popular is how amazingly ubiquitous it is. A fork wouldn't have that advantage, especially in comparison to languages and environments that are objectively better anyway.
Initially, I believe you are correct, but I am not sure long term. PHP is easy for a system admin to install. Other than perl, I am not sure if there is anything out there that is as easy. I would imagine a fork, if it wants to survive, would need to be just as easy.
I do wonder why PHP has never really had any competition in the "super easy for my system admin" category.
I can walk my mother through installation of a PHP script. I can almost walk her through installation of a CGI script in any language, really. But unless that shared host has already set up a Python or Ruby environment (few do; I can count the ones I know on my left hand), I'd never, ever dream of trying to walk her through setting up something in other languages.
Making running scripts as easy as dropping them onto a filesystem would be perhaps the killer feature for other languages to target, if they want to take a bite out of PHP's consumer market share.
Of course, then there's another battle: shared hosting providers. The vast majority are running some sort of third-party control panel software with very opinionated hooks deep into every bit of configuration on the machine. Unless that killer feature can be baked into that software, and unless the shared host actually updates their software, there's no hope.
You wouldn't believe how many shared hosts are running ten+ year old installations of everything, not giving a crap that they're out of date. And hey, that ten year old version of PHP still installed there is still going to work just fine with the majority of scripts...
With their "Move fast, break things" motto I can only shudder at the imagination of the scenarios happening at Facebook HQ:
"Hey, so you are the new developer in town? Can you build service X which service Y and Z can use? Just write a proper API, we need this ASAP."
- "So what do I write this service in? Any framework already build to support these operations?"
"Pick your poison! Check our internal repository collection and see if you can re-use anything useful. As long as we can communicate with it via the API!"
Facebook has their own PHP fork, which they use/and modify. I do not believe that Facebook pushes their modifications to PHP as a whole.
Then they use it only for the view layer, it only pull some data from memcache to render HTML.
But all the hard work is C++ / Java / Tornado / etc
I found a page from 2010, but I couldn't find a more up-to-date source.
http://www.makeuseof.com/tag/facebook-work-nuts-bolts-techno...
- Cassandra (Java) was created by Facebook http://en.wikipedia.org/wiki/Apache_Cassandra
- Giraph (Java) has significant contributions from Facebook and I believe was founded there http://giraph.apache.org/
- FP (Erlang IIRC) at Facebook http://cufp.org/videos/functional-programming-facebook
- Facebook's page listing some of the open source projects they contribute to https://developers.facebook.com/opensource/
The backend is written in everything from Erlang to C++.