Sharing in D

[Walter Bright]

superdan wrote:
> i'm afraid c++ and c++0x and java are "actually fine" to write
> multiprocessing code if the coder is fucking superman on steroids
> after an overdose of ritalin.

There's also the issue of the person who thinks they know how to write 
multithreaded code, but don't. Want an example? Just look at some of my 
older code.

This code may "work" well enough that the problems don't appear until 
years later. Then what?

I remember the reaction to Herb Sutter showing some of the things that 
can go wrong to multithreaded programmers - it was holy ****, I had no idea!

Having no compiler help with this is like turning off static type 
checking. Sure, if you're a good programmer, you don't need static type 
checking!

I believe there is a growing need to be able to statically *prove* 
things about a program. If that can be done, you don't have to rely on 
experts being always right. Java, for example, can prove your program is 
memory safe. Certain classes of bugs cannot happen with Java programs. 
The more classes of bugs that can be defined out of existence, the 
easier it will be to verify correctness regardless of the expertise of 
the programmer.