Not Even My Wife Reads This

I started Colossus in December 1997, almost 11 years ago. Yesterday I turned over leadership of the project to Clemens Katzer. That's a long time to maintain a software project, and it feels weird to me that I finally quit.

Colossus was my first large game project, my first large Java program, my first large GUI program, my first experience working on a complex game AI, and my first time managing a significant multi-programmer software project. Somehow it worked out pretty well, for the first few years. We had a pretty full-featured, stable Titan game, with somewhat dumb but working AI, and support for lots of cool variants. We had a rotating team of up to half a dozen developers at a time working on the game, and it was fun.

But then I added network play, and I didn't do a very good job. At the time there were no Java remote method libraries that had the features we needed, so I wrote everything myself with a simple string socket protocol, with lots of receiver threads. But I wasn't rigorous enough with thread synchronization, which meant walking the line between deadlock and corrupted game data. And gaining acceptable performance meant that a lot of logic that existed on the server side needed to be reproduced on the client side, but excessive coupling in the code base meant this couldn't be done cleanly and a lot of code got duplicated. I tried and tried to fix the mess, but couldn't put Humpty-Dumpty together again.

I finally decided to start over, in a better programming language (Python), with a better networking framework (Twisted), with a better overall design paradigm (game events flowing from the server to the client, using the Observer pattern to reduce coupling and allow complete reuse of the core game logic between the client and server) and with safe sane single-threaded code. Thus was born Slugathon. Unfortunately, as a new father I didn't have nearly as much free time as I did back when I started Colossus, and so Slugathon still isn't finished.

With Slugathon unfinished I felt obligated to keep maintaining Colossus, but I didn't really have my heart in it. I figured my job was to keep the project alive until a successor showed up to take it over. We had several guys on the team who were technically qualified to take over, but none seemed likely to put in the necessary amount of time on a consistent basis. Then Clemens joined, and not only submitted code for his pet features (the fun part that everyone wants to do) but started watching the bug tracker like a hawk, and talking to users about their bugs, and making special test builds so that users could test that their obscure hard-to-reproduce bugs were fixed, and adding documentation of which features went into which releases, and all the other not-fun crap that 90% of small volunteer open source projects can't find anyone to do. Around the same time, longtime contributor Peter Becker did a giant refactoring that reduced the amount of code duplication from disgusting to merely gross. And Clemens switched the server side from listener threads to NIO, reducing the total thread count from insane to merely scary.

Which meant I was no longer needed on Colossus. So I'm committing to releasing a stable, network-playable version of Slugathon by the end of 2008. And hoping that the Colossus team can continue cleaning up the mess I left them. Hopefully someday soon we'll have two stable networked Titan games instead of zero.

(By the way, I played a couple of four-human 2-AI networked Abyssal6 Colossus games this morning. It was great fun, even though both games had technical difficulties.)