On 08/23/2015 02:29 AM, jwsmobile wrote:
There was an outfit here in Orange County, Ca that was a den of
oddballs who actually like to write compilers and were good at that
was called CSPI. In the 78-79 timeframe, Microdata bought a compiler
for Cobol and Fortran from them with a really big check and helped
them along the path of becoming a viable company.
I do remember CSPI.
Anyway, one of the things that Express was competing with were DEC
systems and they benchmarked both the Fortran and the Cobol and
outran DEC. Express was still a miserable failure for them, and they
of course only made the one system design and never enhanced it.
Someone with better knowledge about DEC can guess the 11 and OS.
Anyway another thing they had to do was pass this test, and I suspect
it is the one that Chuck refers to.
If you were selling to the USG, it was definitely in your interest to
pass the Naval COBOL Auditing tests. Grace Hopper was largely
responsible for their development and it stems from a morass of vendor
"extensions and features" and the individual interpretations of the
CODASYL standard that could render compilers different as night and day.
It was apparent that some sort of test of conformity to the standard
was necessary.
FORTRAN suffered from the same problem also. In particular, I recall a
CDC extension that seemed innocent but resulted in ambiguous syntax--a
definite no-no in the world of languages. The PSR was never solved, as
it would have required a change in an existing language.
Note that these tests are not benchmarks--they merely serve to make sure
that the compiler and run-time actually do what they're supposed to--and
not what somebody *thinks* they should do. So these tests are
historically important.
CSPI was bought up by or morphed into one of the outfits which sold
Cobol compilers for the PC, but I don't know which one. If one looks
on bitsavers at the front of the Microdata Express Cobol internal
spec, it was prepared by CSPI, California Software Products, Inc.
I think all major compiler producers had specifications of that
nature--they were also written for operating systems. Paper first, code
later...
--Chuck