1 Oct
2013
1 Oct
'13
2:11 a.m.
El 01/10/2013, a les 0:40, Dave McGuire <mcguire at neurotica.com> va escriure:
As (primarily) UNIX people, we are trained to think in terms of
abstraction layers and common interfaces. The IBM mainframe world in
general, and JCL in particular, find no value at all in that.
Production shops today, on modern hardware, still sometimes think in
terms of allocating cylinders on disk drives for job output.
Not 'sometimes'. Most of times. A IBM educated programmer misses the possibility
of limiting the output storage used by a program. The idea of a runaway program eating a
whole filesystem is... weird. On the other hand, the 'file' concept is abstract in
MVS. The programmer uses 'files' which are bound to real physical datasets via
JCL, so there is some level of abstraction at the I/O level. VMS and RSX 11M+ have the
logical name stuff, which can be seen as similar to the DD cards in a job. Unfortunately
UNIX dies not have anything similar (symlinks being the closest ).
JCL is fast to parse, efficient to execute, and hard to read and
write.
Most of the JCL is generated automatically from skeletons or put together by a program. It
is a little bit hard to code JCL by hand, but it is really easy to write a JCL generator.
The IBM mainframe world never bought into that bullshit idea of
"programmer time is more important than processor time", because, well,
it isn't. The programmer does whatever is fastest for the computer to
execute.
Well, the damn subcapping thing makes an installation to be very aware about CPU usage.
And, one way or another, IBM always wins this war...
-Dave
--
Dave McGuire, AK4HZ
New Kensington, PA