der Mouse  <mouse@Rodents.Montreal.QC.CA> writes:

> > Or perhaps someone could implement the Digital Unix hack of only
> > saving pages used by the kernel -- which would generally result in a
> > much smaller dump.
> 
> It could also lose crucial information, because what you'd really be
> saving is pages the kernel thinks are in use, and its idea may be wrong
> for any of many reasons - indeed, perhaps that's the very bug that
> caused the crash.

That might seem like a good argument, but in fact virtually no bugs
(*none* that I know of) encountered in the last 3-4 years are of that
nature.  And it would be easier to have a knob to change the behavior.

Note that such a hack is also fairly critical for getting *any* useful
information from a system with a large amount of memory.  Disks may be
getting faster, but dumping 4GB still takes quite a while.