larry@JPL-VLSI.ARPA (06/21/86)
The following is Eugene Miya's opinion, not to be confused
with that of his employer or any client company for which he
may be working.
You are right to ask for ("easy updating to future CAISs").
The problem is what modern operating systems are out there that you are
refering to? As you point out EXEC*1100, etc. are not modern OSes.
VMS is modern in my book (with some problems). Unix is not modern,
but a bit better, certainly modular. MACH ONE at CMU? Dave Clarke's New
systems?
Playing devils advocate at that low a level is ridicuous: what about
version control? How many different editors are people going to have to
learn? What about a common debugger? Most systems other than the largest
manufacturers only have Fortran and Ada compilers, so some of the
above concepts are foreign. Try to mesh different files systems and more
importantly higher-level network protocols to transfer data (not
addesses in the CAIS). Frankly at the rate the computer industry is changing,
I have little faith in EXEC*1100, Burroughs MCP, or CDC NOS. Let them die a
quick death rather than a slow protracted one.
Your argument should be based on doing things for one's followers
or one's children. Let's lay the ground work now for our successors than
do a half-assed jobs transitioning from our early mistakes.
Sort of like preventing toxic wastes in ground water. Software
is always difficult to do right the first time, let us not be
too hasty because there is a mid-1987 requirement. What can be a better
low-risk, high-gain? We argued the vendor independent approach and
are doing quite nicely.
From the Rock of Ages Home for Retired Hackers:
--eugene miya
NASA Ames Research Center
eugene@ames-aurora.ARPA