Currently working on an open source implementation of Minecraft [1] - a server, client, and various related things. The new Minecraft update released this week and radically changed how networking works, so it's a bit tough to get it all working again.
It's cool because it's a totally open source recreation of everything. It's got a server, a client, terrain generation, physics simulation, level editing, classic support... this is one of two really big projects I've undertaken.
I'm also still working on MediaCrush [2][3], at least when my other half decides to show up and write some backend code.
> Currently working on an open source implementation of Minecraft
Nice.
Are you aiming to implement a thread-safe plugin API, perchance? The current problem with bukkit (and with vanilla minecraft server, of course) afaik is that whenever a plugin needs access to global state, it uses an API that does not do any locks of any sort; which means that usually, whenever one needs to e.g. write to global state, one dumps these critical parts onto the main thread. Solving this is nontrivial but rather important, I think.
It's just a library, so it doesn't handle plugins. However, the library is thread-safe, and the server [1] I wrote on top of the library supports plugins and is also thread safe. For anything dependent on a specific thread, there's generally a queue to push things into.
That sounds like a very nice and thought-through architecture/design.
What's your current status / how far away are you from a more or less complete beta (e.g. in regards to the minecraft client) (if you don't mind me asking)?
Well, the design has been through a few revisions. I'm not afraid to make breaking changes before version 1.0.0 (for example, I completely rewrote the level handling system to improve level editing performance by an order of magnitude).
The current status of the server is:
* Full support for Minecraft 1.6.4 networking (1.7.x networking is a WIP)
* Full support for editing Minecraft levels
* Implementations of all Minecraft data types (i.e. slots, metadata dictionaries, NBT, etc)
* Slightly buggy emulation of Minecraft physics
* Simple procedural terrain generation
Here's an incomplete list of things that aren't done:
* Lots of block interactions (like water emulation, or falling sand) were done, but were lost during the refactoring I mentioned above
* Redstone and all related logic
* Mob AI (work in progress, not very far along)
* More detailed world generation (missing ores, caves, biomes, villages)
Aside from the server, there's also a mature classic networking library, and a fairly complete client library (no rendering).
The project is more than a year old now and is built on top of code that's about three years old. It's a pretty big undertaking, and it's going to take a while to finish. It'll be done faster if I get more third party help, though!
Thanks for your detailed response. I see quite a few people are involved, and the project is a truly active one; I wish you luck. :)
One last inquiry: have you done any benchmarking / performance tests on your code? I imagine that since the library is thread-safe, it should in theory be able to handle significantly larger numbers of online users, etc.?
It's been a while since I've formally tested it, but it performs very well. I don't have hard numbers handy, but CPU usage and memory usage is both well-tuned (and better than vanilla/bukkit). I've tested with a few dozen concurrent players and it didn't have any trouble. All players reported smooth gameplay, far better than vanilla or bukkit.
It's cool because it's a totally open source recreation of everything. It's got a server, a client, terrain generation, physics simulation, level editing, classic support... this is one of two really big projects I've undertaken.
I'm also still working on MediaCrush [2][3], at least when my other half decides to show up and write some backend code.
[1] https://github.com/SirCmpwn/Craft.Net
[2] https://mediacru.sh
[3] https://github.com/MediaCrush/MediaCrush