Here's an update on where things stand - there's been considerable progress over the past two months.
Colin McLean has joined the project here at Penn, and has started to write kernel code. While he's just getting started on coding, he's coming up to speed quickly, and I expect will soon be a major contributor. Other people are starting to show interest in contributing as well.
We now have two PCs here in the lab that are being used for EROS work, in addition to my lab at home.
But perhaps the most exciting news to all of you is that the EROS kernel is once again beginning to limp along. After a while of catching up with the changes in linux (our development platform), the bootstrap code is once again working. The kernel gets loaded, successfully establishes a virtual map and an interrupt environment for itself, and halts just prior to running the autoconfiguration logic. In spite of how it may sound, getting this far is a significant accomplishment. We will in short order be on our way to running user-level code off of a ramdisk, and I expect that Colin will be starting to build disk drivers any time now.
On another note, the topic of whether to make the kernel multithreaded or not has come up several times on the dimsum mailing list. We resolved in the discussion that we knew how, but it wasn't clear whether it was worth the bother. In light of some discussions with Jeff Chung here in DSL, I've now committed us to a threaded kernel architecture. Drivers will have their own threads, and will be scheduled and prioritized using the same mechanisms as user-level code. These threads will, however, run as "parasites" in the address space of whatever process is handy, making them lightweight..
That's all for now. We'll send out updates periodically.
Jonathan