Sunday, July 22, 2007

Erlang-Style Concurrency in Google Gears

I heard about Google Gears when it came out a few weeks ago, but only a couple of days I took it for a short test drive. I was surprised to see that Google Gears uses Erlang-style concurrency (well, sort of).

Google Gears supports asynchronous background processing using the WorkerPool module, which lets your Google Gears app execute multiple tasks concurrently without blocking the main UI thread. What I find interesting thing about WorkerPool is that threads only communicate by message passing and that there is no API for synchronizing access to objects that are passed around between threads. This is a departure from the traditional concurrency toolbox in imperative languages -- using synchronized access to shared memory.

It's so easy to use it's almost... Erlangy :)

So how does Google Gears pulls off this feat of simplicity?

Google Gears only lets you send strings between threads. Since Javascript strings are immutable, there's no danger in letting multiple threads share them.

This solution is clever, but it comes at a cost -- you must serialize your objects before sending them between threads, which incurs a performance penalty (some extra coding, too). I don't expect the performance penalty to noticeably impact most Google Gears apps, though, which are mostly offline adaptations of standard webapps.

Google Gears has managed to pull off easy concurrency in an imperative language by relying on the immutability of strings in Javascript, but this approach is only practical on the desktop. In a server app with higher performance and scalability requirements, sending serialized objects would probably slow things down too much.

All in all, I think Google Gears is a great solution for adding offline support to webapps, and it gets bonus points for using Erlang-style concurrency. Now I'm just waiting for Safari support :)

1 comment:

Erlang, or Utility-computing vs. Appliance-computing « X marks reality said...

[...] built upon Erlang. IMDB (owned by Amazon) is switching from Perl to Erlang. Google Gears is using Erlang-style concurrency, and the list goes [...]