> 
> How, exactly, would we replace sbrk with mmap()?  By keeping a s3kr1t
> global that's the highest address that has ever been returned and
> mapping there with MAP_FIXED?  There are myriad other icky issues.
> 
Main question: Must the sbrk() call be still around? I doubt there are
many programs still around today that use sbrk() directly; if they 
do I should more think of considering them broken.

The best thing to do would be to just remove sbrk() and rehack malloc() 
to use mmap() instead. This would also solve the problem with giant
page tables on e.g. vax :-)

Would anyone complain much if sbrk() et al were removed?

-- Ragge