> I parallelized the SetL runtime for shared memory multiprocessors, with dynamic load balancing, parallel GC, and added explicit parallel constructs, at Urbana

> I parallelized the SetL runtime for shared memory multiprocessors, with dynamic load balancing, parallel GC, and added explicit parallel constructs, at Urbana-Champaign, from fall 1988 over the next couple years. It's weird, to me, that the global interpreter lock persists in Python, 3 decades later. I don't think anyone thinks GvR is incapable of removing the GIL. Given the comparatively tiny base of installed SETL software, surely it's a lot easier to check that any backwards-incompatible changes to SETL won't break any important installed software based on it than it is to do the same for Python? (And, as a follow-up, I think it's fair to say that there is no such thing as a fully backwards compatible change. My background is with Perl more than Python, where the problem is especially acute; but I imagine with Python too, it is very hard to make a change where one can be totally sure that the behaviour being 'fixed' isn't behaviour that some piece of software somewhere out there relies on being broken in exactly the historically right way.)