It's a bit too early at this point. I've saved a ROM dump from my camera, just in case. Overall it works pretty well, but it doesn't feel as solid as it's on 5D2 (I've got some random ERR02 messages).
I'm thinking to start with a beta with few features (for example, start with zebras, peaking, histograms), have a few early testers run it for one week, and if it's rock solid, I can add some more things. Then repeat this process as many weeks as needed, until the port is complete.
There is an issue that's stopping me for publishing it: how to handle multiple cards. Right now it's hardcoded to work on CF cards only. As long as you don't use any features that access the filesystem (like movie logging, HDR post scripts or other files created by ML), you can also boot ML from a CF card and save pictures/movies on a SD card.
I could compile two binaries - one for CF cards and another for SD cards - and select one automatically. I have to find a way to detect which cards are loaded from bootloader stage. I'm afraid that letting the users manually choose the right version will create trouble.
Restoring ML after format doesn't work, so probably the best way to deploy it will be to load ML on a SD card (which will always be in the camera), and shoot on CF cards. Or viceversa (load ML on a CF cards and shoot on SD cards).
Handling multiple cards requires some major code changes, and I've already had trouble with the most basic file I/O calls (
http://www.magiclantern.fm/forum/index.php?topic=2528.0 ) .
Since my proposal of limiting it to CF cards was not welcome, I think we'll have to wait until these things are solved.