Yes, absolutely. A project I've wanted to run with for a bit is writing a Scheme implementation in PyPy. It would provide a wonderful way to better learn the parts of Scheme I don't know, and use Scheme alongside Python code (which is nice, because it would mean I could conveniently use Sexps as a data language and do the data handling in Scheme, but do the real work in Python, for example). I'd been looking into some of it idly, but as you say, there isn't much.
More significantly, one of the projects I've been working on really requires a new language, and while I could implement using, say, PyParsing and so on, having free JIT is really nice.
I've actually been working on a Scheme to Python bridge, but it's currently very primitive and exists mainly to allow embedding the Python interpreter in PLT Scheme.
I'd love to see this. Python just has better libraries in the most important areas: numerical computation, machine learning, etc., than any other platform. It's main downside is that you can only get at them from one language.
One thing I would be interested in hearing that I haven't heard anyone mention is how far it seems like a language can reasonably stray from Python without essentially hitting its head on PyPy at every turn. For example, functional languages have a hard time on the JVM due to its lack of tail call elimination.
Most definitely. It would be interesting to see how this compares to using parrot as your platform. The two approaches differ completely in ideology, and I wonder what would be easier to develop on.
Yes, very much so. I'm currently designing and implementing a language in Python (designing and implementing it together, to see what features work). I'm considering eventually implementing it on the JVM, to take advantage of Java's multitude of libraries, but PyPy might be another option.
More significantly, one of the projects I've been working on really requires a new language, and while I could implement using, say, PyParsing and so on, having free JIT is really nice.