MLton RAM usage

[Stephen Weeks]

> > Unfortunately, that doesn't work.  For example, on my machine, here is
> > /proc/meminfo.
> > 
> > % cat /proc/meminfo
> >         total:    used:    free:  shared: buffers:  cached:
> > Mem:  525217792 522846208  2371584        0 99725312 276832256
...
> Well, it looks like you've spent all your memory - how would it
> ever come to run like this?

The problem is that I haven't spent all my memory.  In fact, all I have running
is X, xemacs, ical, and netscape.  I'm no kernel expert, but I'd say the problem
lies in the 375M of buffers + cached which is available to run programs if I
want to.  Henry can probably fill in more details on why free is such a bad
underestimate of what's available.

> Oh, and another thought - what if a user is running two
> different apps which are both compiled with MLton? :)

I agree this could be a problem, but recall that the memory limit that we're
talking about is only grabbed by a MLton app if it needs memory.  The compiler
is such hog that it often needs it, but many other apps don't.

I don't disagree that what we're doing now could lead to problems.  It's just
that I don't know of a better solution than our current over-approximation + the
ability to manually limit memory usage.