I remember when structured programming came out and many in-line Cobol
programmers told me, and correctly I might add, that structured programming
runs slower than inline programming. I countered that "yes, it will run
slower but it will be easier to maintain". MPE faces a similar challenge.
If we tie each version of MPE to a particular piece of hardware, which will
run better, we too will get stuck on the treadmill that CSY is currently
running, and we too will run out of resources. Having a full version of MPE
for each type of hardware seems to be an impossible task since even HP
cannot maintain a version of MPE for each PA-RISC processor. Only the
government can afford to buy a specialized hammer for each particular
situation. The economics in play today will not support such
specialization, IMHO.
If I understand Wirt correctly, the purpose of emulation is to get us to new
hardware quickly and to provide the legendary backward compatibility for
which MPE is known. I don't want to run everything in an emulation mode
forever but it seems to be a necessary bridge between the past and the
future. HP understood this in the conversion to Spectrum. They wrote an
emulation layer that is still in use today. If IA-64 is the future, why
spend time on PA-RISC?
Regards,
Mark
-----Original Message-----
One of the areas that I would be concerned about is that MPE has
historically been very good, partially, because of it's tight integration at
the hardware level. I personally am one of those people that does not
believe there is a single programming language, O/S, or hardware solution
that is best for every single situation. Each one has its strengths and
weaknesses. If you want the best solution for a particular problem you use
the tool that is best suited for the problem. Emulation at a certain level
would have to come pretty close to making it seem like I had done this at
all levels.
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|