Patrick Logan heard Maurice Herlihy give a talk, and elevated him into the pantheon, which he richly deserves. I know him mostly from his work on wait-free data structures, which predated his arrival at Brown. If memory serves me, we overlapped just a little bit there, but I don't remember having any significant interactions with him.

Patrick then went on to extend our VM discussion by pointing out that

... this level of concurrency control should at least show up at the virtual machine level. Why emulate a 1970's instruction set?

Now there's a thought. Definitely worth a master's degree project. Might be worth more if things turn out hairy. And be worth a lot more if it actually turned up in Parrot or Mono. Truly told, there are a ton of old (and good) research results sitting on shelves in university libraries. Why aren't we picking this low hanging fruit in the open source world?