In <[email protected]>, on 03/13/2013
at 09:47 AM, Gabe Goldberg <[email protected]> said:
>I'm writing about "Back to the Future" for mainframers -- historic
>(but sometimes forgotten) mainframe lessons needed by and
>best-to-be learned by new mainframers (but everyone, really).
Most of this is generic. In no particular order:
Test your backups.
Have a plan B.
Ask questions. That includes asking yourself what you really
mean.
Question your assumptions.
Document first, then keep your documentation up to date.
Maintain a suite of regression tests.
Many one-time tools aren't.
Never throw away test scaffolding.
Don't update the running system.
Have a migration plan, and include fallback as part of it.
Be wary of local optimization.
Understand that in a long sequence of trials the probability
that some[1] rare event will occur approaches unity.
Ask for needed approvals, changes and resources well ahead
of time.
Use standard system services rather than rolling your own.
Write maintainable code.
Code and design reviews can help to circumvent blind spots.
Alway distinguish between the symptoms that you actually see and
what you believe is causing them.
Be paranoid. Assume that your data are dirty and that there will
be future changes[2] in the specs.
Read the reference manuals for details; don't trust 2rd party
texts or your memory of previous releases.
Preserve your source code, including old versions.
Keep production software under CM control.
Don't code "magic numbers"; define named constants and equates
instead.
Use vendor-provided mappings rather than rolling your own.
A few coding techniques for newbies to learn:
The use of UNPK and TR for converting to hexadecimal.
The use of BXH or BXLE for testing successive bits in a string.
[1] There's no way no know *which* rare event; plan for all
that you can forsee.
[2] From Datamation
Then he turned with a snarl
and he said with a taunt
it's just what I asked for
but not what I want.
--
Shmuel (Seymour J.) Metz, SysProg and JOAT
Atid/2 <http://patriot.net/~shmuel>
We don't care. We don't have to care, we're Congress.
(S877: The Shut up and Eat Your spam act of 2003)
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
