Marco.org

A big win for Facebook’s alternative PHP engine.

I ran Overcast on HHVM for a while last summer, but there were still too many bugs and incompatibilities for my comfort. There’s still a while to go before it’s a drop-in replacement for everyone — as you can see, Wikipedia required a lot of custom work from HHVM’s developers — but it’s making progress.

Still, I question whether alternative runtimes and custom language extensions for a very old, limited, and flawed language are worth the trouble, incompatibilities, fragmentation, and potential instability for sites with fewer engineers and less of a PHP investment than Facebook and Wikipedia.

Having experimented with Node and Go over the last two weeks, I can’t help but think that most PHP developers and sysadmins would be better off either sticking with official PHP, because it’s much easier and safer to rely on, or beginning a migration to a more widely supported new language that’s likely to have a longer useful lifespan.

A large investment in HHVM or Hack today seems to be worthwhile only if you’re paying much more for PHP CPU usage than engineering salaries, you have a lot of unused engineering capacity to migrate to and maintain the new setup (which takes more effort than official PHP because it’s so bleeding-edge), and your codebase is so large that it’s infeasible to start migrating away from PHP.