[mmckeen]

Knowing how large Facebook's PHP/Hack codebase is/was, rewriting that application would have been a disaster.

Hack is a fantastic language and the work by the HHVM team benefits the PHP ecosystem as a whole.

PHP may have its problems, but many companies use it for its many benefits, mainly the ease of finding developer talent.

[fishnchips]

> rewriting that application would have been a disaster

Or the best decision ever made by the company.

[nkozyra]

Based on what? They rewrote the language into another that is largely backwards compatible, fixes PHP's quirks and is insanely fast.

Given the choice of migrating / transpiling an entire codebase or making the underlying interpreter / compiler and language fantastic and speedy, the latter is a big, big win.

What would rewriting the whole thing in, say, Ruby or Python have done beyond limiting the amount of "PHP sucks!" critique they so often get?

Seriously, take a look at Hack. It's very well done.

[fishnchips]

This is not to diss any particular language but an architectural problem - frontend and backend are very tightly coupled which makes it difficult to develop anything outside this stack. By splitting a humongous (and always growing) homogenous application into smaller services you get smaller failure domains and better interoperability with acquisitions and potential new groundbreaking projects that do not easily fit into the current development model. Again, I have no opinion about HHVM and Hack which are completely orthogonal to the problem at hand.

[nkozyra]

Can you expound on this? I don't perceive front end and back end being especially more coupled than, say, Node.

[fishnchips]

The coupling has nothing to do with the language and everything to do with the fact that FB grew organically into this giant monolithic mess of spaghetti. Can you do SOA in PHP? Of course. Does FB do it? Nope. Will this bite them in the long run? I believe so.