On Tuesday 11 July 2006 21:09, Audrey Tang wrote: > I really cannot argue with that argument (essentially "let's punt and > see"); therefore this ticket is probably best reserved until Parrot actually > has a security model, in which time I'll then argue that :immediate should > be subjected to severe restrictions.
Why? Because people might write code, by hand, that does careless things in :immediate subs? People can write HLL code that does careless things in :init subs too. The only difference that I can see is when those careless things happen -- whether at compile time (when creating a PBC file) or at run time (when evaluating PIR or PBC). Any computer language that allows people to do the programmatic equivalent of running and allows people to do the programmatic equivalent of holding scissors likely allows people to run with scissors. Now you can disallow running or holding scissors, but I'm not sure it's possible to disallow all potentially dangerous combinations with static analysis. Thus my suggestion is to write, in bright friendly letters, "People who do stupid things often die." and let bitter experience underline that sentence in the minds of people who refuse to believe it. Not that I've *ever* written code that deliberately subverts static analysis, oh no... certainly not to prove a point. -- c
