After a few years of observation, I reached a disturbing conclusion. I'd like nothing more than you to prove me wrong.
There are Open Source games of impressive quality. Especially no original games.
There are lots of clones, though. Freeciv may be good (I don't like this kind of games except for Master of Magic), but my friend - who likes civ-like games a lot - tells me it's it's little more than well-made Civilisation clone. I asked him if Freeciv has anything that hasn't been done before in commercial Civs. He says there isn't anything like that. Freeciv may have a little better network support, and that's it.
Some people like Wesnoth, but I don't. I really don't like the graphic style, and the game is too simple for my tastes. Even if I liked it, it's been done before.
To my knowledge, there's no good opensource RPG game, most probably because rpg games require a LOT of time to build convincing story and dress it up in pretty clothes.
TA Spring doesn't even try to pretend it's not Total Annihilation clone, even worse - it requires a lot of commercial data to work, and free media replacements for the game are far from working.
Freecraft - the name says it all.
Attal - Homm clone. Freelords - ...
So how about fps games ? Cube/sauerkraten may have its advantages, but quake-style deathmatch has been done to death.
Best opensource fps games are based on ID Software engines, not that I don't appreciate the fact ID has a tradition to regularly release outdated engines. The point is: I know no original opensource game based on them. Mostly quake-style deathmach clones.
Tremulous is fresh enough, but suffers a lot from bad design and lack of punkbuster.
Arcade - clones, clones everywhere.
There are some truly fresh opensource games, but they're few and far between. I can think of 3:
Liquid War - althrough fun, it doesn't last long. (available in ubuntu repositories)
DROD - in my opinion the best puzzle game ever made.... but it didn't start as opensource game.
By the way: not all people know the last DROD game (JTRH) is actually GPLed despite being commercial. They just charge for extra graphics and official levelset. You can still play usermade levels (LOTS of them), and use fully functional level editor !
Linley's Dungeon Crawl (http://dungeoncrawl.org , but also available in ubuntu repositories)
In my opinion the best, most polished, varied, balanced, and fun roguelike ever made. The only thing it lacks is good documentation. But again, it didn't start as an opensource game. A.D.O.M. doesn't even come close.
So, is opensource even remotely viable way of developing games ? I'd love to say 'yes!', but my experience says otherwise. Prove me wrong.
I don't get one thing. All it takes to make a good game is a brilliant, fresh idea. (unless you'd like to develop a RTS or FPS game; these require very efficient and advanced code). Master of Magic is played to this day by some hardcore fans (dosbox; the game available from the underdogs), despite having truckloads of bugs !
Are fresh game ideas so hard to come by in opensource world ? Can't be !
One thing that personally discourages me from contributing ideas to opensource games (or any games for that matter) is that usually you get no credit whatsover. Many of my suggestions have been implemented in games, and I'm only credited in one (Lbreakout2 uses many of my powerup/brick ideas). On the second thought, there's a mention of 'b0rsuk' somewhere in Wesnoth changelog, too.
(the author of this rant would love to make some games himself, but he has a long way to go from being just IT student/linux fan. He'll be making games all by himself, because you can't count on others.)