I am not sure if I am in a position to comment here but it does seem that
some proposed features in MRI may in essence make it more "static" and thus
a harder platform to evolve for the future.

AST inspection- I am not sure what the exact motivation is for this but it
does expose portions of the frontend of the compiler system which basically
adds to the "spec" of the language.

YARV bytecode- I think the same comment applies to this- and I am uncertain
if the YARV byte code system effective as it has been as an internal
representation for the 1.9 releases would be as an exposed portion of the
spec. My "gut" tells me that in a JIT type situation certain opcodes could
probably be removed and be inferred by the system. At an extreme would be
something like the high level byte codes used by Craig Chambers.

Just my .2 cents,

Carter.