Wayne writes:
> Jeff: This is a key issue.  No source code equals no
> enhancements to MPE in the future.

This really isn't true, because once you have MPE running on top of a
platform emulator there aren't that many enhancements that would
actually require changing the MPE kernel itself.  Almost everything can
be done either as ordinary user-level and PM add-on software on top of
MPE, or as changes to the emulation layer *under* MPE.

SO all new hardware platform and device support becomes an emulator
issue rather than an MPE issue.  The emulator simply needs to make your
new 10TB disk drive look like a lot of smaller disk drives that MPE
knows about.  Your super ultrium tape drive can be made to look like a
DLT4000 or something like that, etc.  As long as any new hardware
peripheral can meaningfully be thought of as a class of devices that MPE
already knows about (Tape, Printer, Disk, Network, etc.) then the
emulator layer can dynamically translate command sets, interfaces, and
so forth so that MPE never knows what's happening.

Most functionality enhancements that MPE might need can be done with
ordinary programming on top of the OS just as they could be done today.
You might not get quite as integrated of a solution as if it were done
in the kernel, but it will work fine.

Facilities like the AIF Procedure Exits can be used to intercept MPE
operation at various places to implement things like new file types, new
CI commands and so forth, again without the need for access to MPE
source.

About the only kind of change which would require source changes in this
environment would be capacity increases like the maximum number of
processes or users.  For these, we're likely to be stuck with the 7.5
values (and whatever HP does in this area in the next year) forever.
But the current limits are larger than any current customer needs (so
far as I know) and there are still other alternatives (run multiple
instances of the emulator on a system and use NetBase or something to
link them), so I don't see MPE capacity limits as something that's going
to be a big problem.

This is perhaps the most important feature of the Platform Emulator
concept.  It actually allows MPE to move forward on supported hardware
platforms and allows it to evolve dynamically to unknown future changes
in the computing environment.  So if today's TCP/IP networking is some
day insufficient, an emulator will be able to dynamically translate
between TCP and the flavor-of-the-decade (or century) whereas the old
"real" 3000 box in the corner may actually be useless at that point.

I'd like to know what sorts of "enhancements" people feel would be
important to have in MPE going forward in order to make them feel good
about continuing to run on it.  My guess is that the Platform Emulator
would be able to supply pretty much all of them.

Gavin