[Cryptography] GCC bug 30475 (was Re: bounded pointers in C)

[wish at dumain.com]

Thus Spake Bear <bear at sonic.net>:
> Given the history of bugs, I would rather have a language be 
> cripplingly slow on some hardware (doing all math via emulation, 
> yuck!) than have important parts of mathematical semantics 
> unknown, unknowable, or unenforceable on all hardware.  If I 
> want to use Java on a 1's complement machine, performance will 
> suck but it'll have no math-related bugs that it didn't have 
> on the original.  

If you want widespread adoption highly visible slowdown is more
imporant than largely invisible until it bites you in the ass safety.
Of course you only really need to optimise your language for x86(_64)
and ARM(64).  

'C' is designed for the 'give the programmer enough rope to hang himself'
crowd who want to trade safety for speed on the assumption that the
programmer knows what they are doing.  I think trying to convert 'C' to a
different philosophy is unlikely to succeed as a lot of people buy into
it.  It would be more productive to switch to a language which has more
of a safety orientation in the first place and add whatever is needed.
Ada with the Ravenscar profile seems fairly close, has free,portable
compilers and AFAICT can be called from other languages fairly easily.
You could use a conservative 'C' to Ada translator to switch existing
code.  Anything that can't be trivially translated probably needs 
attention anyway.  

William