Monday, January 22, 2007

What was I thinking?

Late 2006, I posted a rant on the state of concurrent processing on Python.

Looking back, I don't really think I was thinking very clearly that day. I imagine I was contemplating a multi-threaded raytracer in Python. A fun idea, but kind of pointless really. Stupid even.

Since that post, I've implemented exactly one program where I had to explicitly use threads. Why did I need threads? Simply because I wanted non-blocking IO, and couldn't be bothered using other asynchronous mechanisms.

The lesson? YAGNI.

2 comments:

Anonymous said...

> I ... couldn't be bothered using
> other asynchronous mechanisms.

That's a pity. Sure, at first it's not easy wrapping one's head around Twisted Deferreds, but once you do, you feel peculiarly... empowered. ;-)

There's a donwside to async code: the inside-out appearance it takes on. That can now be minimized, thanks to Python 2.5, Twisted 2.5, and the new inline callbacks.

Give it a spin, you'll not regret it.

Steve said...

Of course there are some users who really do need concurrency across multiple processors, and they are maybe beginning to wonder about their migration path as multicore processors and multiprocessor architectures become more common.

So, while it may be true that you aren't going to need it, that doesn't mean that nobody is going to need it. The stackless environment with its ability to pickle tasklets on one machine and unpickle and activate them on another looks like a potentially usable solution. There may be others.

Popular Posts