Re: How about a Safe Virtual Machine?

Stephen D Crocker (crocker@tis.com)
Sat, 1 Oct 1994 03:06:32 +0100

I haven't looked at Safe-Tcl, but I doubt there's any non-linear
performance penalty. If Safe-Tcl is Turing-complete, you can build an
interpreter in it for any language you like. There'll be one
interpretation cycle in Safe-Tcl for each atomic action of the
interpreted language. If there's an upper bound on how long it takes
to interpret all possible atomic actions, then that's the upper bound
of the coefficient of the linear relationship.

Unfortunately, it's not enough to limit the time penalty to merely
linear. People go to great extremes to save relatively small factors
-- at least small by theoretical computer science standards. Even a
modest factor of two seems interesting to people, to say nothing of a
factor of ten.

Steve