DR>
  > JdeBP> BTW, I regard this talk of "only 512Mb" as
rather unrealistic.  The
  > JdeBP> people who usually use this phrase forget that any process that
  > JdeBP> required this much memory on *today's machines* would be needing ove
  > JdeBP> 480Mb free hard disc space for the swap file.
  >
  > You're confusing physical memory with virtual memory.
DR>

  Nope.  If a process requires 512Mb of virtual memory on a system with
  32Mb physical memory, it will need 480Mb swapfile.

DR>
  > instance, a database manager can allocate a 2 Gig virtual buffer
  > but not commit it.
DR>

  A specious argument, IMO.  If you will not ever need to commit all of
  the 2Gb, then don't allocate the address space, and only allocate the
  maximum that you will ever need.

  Otherwise you are wasting an large amount of system resources on empty
  page tables (think of what happens in the page replacement and page
  aging algorithms, for one thing).

DR>
  > It is not necessary to have enough physical memory to actuall map the
  > whole file at once, but it is necessary to have enough to
  > virtually represent the full file in the process' address space.
DR>

  A similarly flawed argument.

  On the one hand, in the limiting case the file will be 2^32 - 1 bytes
  long, taking up the whole of the virtual address space for the process
  and leaving 1 byte for code, data, and stack.

  On the other hand (in case you were about to say "but no file will
  ever be that large"), if memory mapped files are to be limited to a
  "reasonable" size, I'd still say that 512Mb *is* a pretty reasonable
  size, especially as it is larger than a lot of people's total hard
  disc.

  As I said before, down the road there may come a time when the 512Mb
  for each individual process limit may seem a little tight.  But that
  is not for some years yet, and by the time it is, the requirement for
  16-bit compatibility and tiled addresses in OS/2 will be long gone.
  So anybody getting worked up over "only 512Mb" on *today's hardware*
  needs a reality check.

  Incidentally, out of interest, OS/2 for the PowerPC does not include
  any such 16-bit legacies.

  > JdeBP <
___
 X MegaMail 2.10 #0:

--- Maximus/2 2.01wb