It sure is good to 'hear' that Chris is still around and still interested in
HP3000 matters.

I'd also like to submit a vote of agreement on his comments about REX.
Initially I was reluctant in accepting REX out of concern that the generated
code would bring our 'overloaded' Series 48 machine to its knees. I quickly
became a convert and have used it ever since.  In fact I wrote a program to
perform a mass load to an Image database about half an hour ago.

At one point we had REX on all 3 of our HP3000 and had developed a number of
applications with hundreds
programs using REX.

Even today REX is used so widely in our site, that we decided to license the
source code from DRA.  We made the support routines y2k compliant and enhanced
the date format, 'DF=', to handle 4 digit date formats.  The original intent was
to get some help from the folks that wrote 'SPLash' and be able to generate
native mode executables but those folks have been too busy with y2k issues and
that aspect of the project has not moved much.

I keep in touch with a number of programmers that have moved on to other
companies and every one of them
indicates how easier their programming assignment would have been if  they had
access to REX.

Regards
Paul Christidis




I used to work on the REX project, and for a number of years
actually maintained and enhanced the compiler. If you made a support call during
a 10 year period, you had a good chance of talking to me.

REX was a Pascal program based on the original Zurich P4 Pascal compiler (if my
fading brain cells remember it right.) It produced basically
assembler
code, but as there was no assembler for the classic HP3000s, it produced SPL
code with lots of assemble statements.

When the HP PA machines came out, a project was started to change it to
produce
U-code (HP's name for basically what other people called P-code.) I even
attended
a special traning class on using HP U-code.  At the time HP was going to
allow
a small group of vendord access to it so they could get native mode speed
and
functionality. Generating U-code meant the HPs optimizers would work, and
that changes
in underlying machine architecture would automatically be used as well.

Suddenly and un-named manager at HP killed this U-code project. I never
heard a reason,
but guess they decided it was just too much trouble. We talked about trying
to then
use Splash, but everyone seemed to lose interest at this time. It still
functioned ok
in compatability mode but was limited by stack space.

REX was the best language I ever used, and I often wished I could have
ported it
to Splash and made it freeware. I still find myself writing Cobol programs
and wishing
I had REX so I could develop it in half the time, and have a much clearer
program.
REX was sort of a reduced intruction set language. It could do everything
Cobol could,
but was much faster to learn. Since it was HP3000 and Image focused, typical
programs
were much easier to develop. It had a little in common with Transact, but I
thought
it was much cleaner.