On Mon, 22 Jul 2013 16:51:51 -0500, Tom Marchant wrote:> 
>>Where can I find some Rules of Thumb; an SMP/E style checker?
>
>I'm not aware of either.  There is the Packaging Rules manual, but there is 
>much that 
>is not covered there.  It is focused primarily on correct SYSMOD construction. 
> Some 
>other considerations for what I would call proper SMP/E packaging include:
>
This briefly mentions tape construction, but makes no mention
of packaging for network delivery.  I submitted an RCF on this
a couple years ago, which was received favorably, but no action
is evident.  I'm awaiting the z/OS 2.1 version with bated breath.

>o Providing useful system holds as necessary
>
We try to.  Feedback welcome.  If any customer has attempted
such feedback that hasn't reached me, try again.

>o Using Fix Categories and SOURCEID to simplify maintenance
>
We don't.  I believe Fix Categories is new; we haven't caught up.
What's needed for SOURCEID?  Would partitioning according to
our monthly regression test cycle be of value?

>o Providing frequently updated downloadable Enhanced Hold data for errors
We provide HOLDDATA; not Enhanced.  But a customer has complained
in this list that while timely it's poorly organized.  We use IBM's HOLDSYS
categories.

>o Proper SUPs for error holds when a PTF resolves the error
>
We do.  FSVO "proper".  Through 3 corporate ownerships we have
never had a problem tracking system compatible with IBM's
naming conventions, so our resolving PTFs SUP the PE SYSMOD ID,
not an incident ID.  This is legal according to SMP/E (although I
have repeatedly needed to submit RCFs whenever an SMP/E
manual inadvertently asserts otherwise.)  I don't know how this
is likely to work if a level-set PTF both SUPs a SYSMOD ID and
uses that in an RMID operand of an element.  But we've never
generated such a level-set PTF.

>o Having a data base of APARs and PTFs that customers can search
>
Don't know.

>o Cross-product dependencies (IF...REQ) where applicable
>
We try.

>o It should NEVER be recommended that a customer use BYPASS(ID) or BYPASS(REQ)
>
Our policy.  Violated once by a rogue tech support person, with
disastrous results.  We have since reinforced the policy.

>o Network delivery for products and service
>
We're a small appendage in a large corporation.  Our downloadable
service is simply SYSMODs with inline elements, zipped according to
corporate standard.  Nothing similar to ShopZ.  For similar reasons
our downloadable products are SMP/E RECEIVE FROMNETWORK
file hierarchies, all zipped.  If we tried to come closer to IBM's
conventions, I imagine a network wonk's asking, "FTP!?  What do
you want to use that old thing for?"

>The products and maintenance should be thoroughly tested.  In order for the 
>testing to be 
>thorough, many permutations need to be tested.  Most PTFs should be able to be 
>applied, 
>restored and applied again without issues.
>
I believe no vendor is capable of testing, e.g., all possible combinations
of ten PTFs with no declared interactions (Cartesian product: 2**10).

For major PTFs (SPEs), RESTORE is usually tested collaterally to
iterative development.  Minor PE PTFs are likely to be tested and
go to field with no RESTORE tested.  This has not been a problem.
Except...:

We deal with an ISV cross-compiler vendor who doesn't understand
the mechanism of SMP/E RESTORE processing.  I believe we've
circumvented most of the problems they've introduced on this account.
But I need to sit down and talk with them at length about the
difficulties they're causing.

>This is not likely an exhaustive list.
>
Any comments on post-APPLY link edit and script processing?
We don't do this; it might simplify some processing at the
cost of weaker control.

Any comments on tailoring of data set names and attributes?
The z/OS 2.1 symbol processing in SYSIN data sets will be a
godsend for this, but only after all our customers have migrated
to a z/OS level that provides such support.

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to