James Mastros <[EMAIL PROTECTED]> wrote: > Dan Sugalski wrote: >> >> Yeah. We're ultimately going to have to add signals to memory allocation >> and IO as things that the embedding environment controls.
> Is it just me, or does this sound awful similar to the existing list of > platform-dependent code? Does this suggest that we should be looking at > having the platform functions pluggable at runtime? Why? The configure system puts together the necessary platform pieces and make generates finally a parrot executable for this very platform. You can't transfer executables from one system to another and hope that it will run (or yes you can copy them, but ...). > just making them a full-blown ParrotSystem object out of them, though > that may raise "interesting" circular problems.) A ParrotSystem object OTOH isn't the worst thing. It could provide the namespace for all possible system functions without the need for numerous new opcodes. > -=- James Mastros leo