Agreed that PCCMETHODs need to be optimized, and they will be. (One of
the first optimizations will be making it so a PCCMETHOD call doesn't
have to check whether it's calling an old METHOD.)
But the old METHODs aren't sufficient to support the dynamic languages
we're targeting. They don't support the Parrot calling conventions that
are now standard. They force rather limited interfaces on several
essential core PMCs, as well as an unfortunate divide between the
features possible in low-level (C) and high-level (PIR) classes.
Optimizations are important, but not at the cost of critical features.
Allison
