O(n^2) is bad - can it be fixed?

[Alex Martelli]

"Tim Peters" <tim.one at home.com> wrote in message
news:mailman.990866978.26483.python-list at python.org...
    ...
> > then I agree that it is so hopelessly broken that Python shouldn't
> > deal with it at all.  Yes, at all.  Python, as an application, don't
> > have the VM tools to do this efficiently.
>
> It could, but, again, it's absurd:  Python could do its own heap mgmt on
> Win32, using the Win32 API heap mgmt functions instead of MS's libc's.
> They're adequate to the task.  But I'm not adding gobs of delicate
> platform-specific code to worm around a problem nobody has <0.9 wink>.

As I recall from back when I was focusing on heap fragmentation
issues for our apps, it's not libc's fault (not in VC++6.0 at least,
but I think I recall 5.0's was decent already) -- recoding at a lower
level left gobs of problems in Win95 (and I believe Win98 did not
change much there).  Using NT4 made the problems go away (even
with libc), so that's what we advised to our customers, and it happens
to be MS's party line too -- "forget Win9x, that's just for games, NT
is the OS for professional use".  Maybe XP, allegedly based on the
NT/2000 underlying kernel technology, WILL solve things...


Alex