On Fri, 17 Jun 2016 14:39:24 -0700, retired mainframer wrote:
>Won't that break at least one of the CBT utilities that resets or extends
>the PDS directory by writing to it directly?
>
And the world would be better for it. Think of all the resource spent
in data centers by programmers recovering from overwritten directories
and corrupted attributes and by contributors to this list attempting
(often misguidedly) to help them, which could better be spent as cash
to buy more IBM products.
Alternatives:
o Use PDSE
o Re-allocate and copy.
Perhaps there should be a new DD keyword, SAFE={YES|NO} where
"NO" would suppress as far as possible any error checking not needed
to protect VTOC integrity. And IBM support could turn a deaf ear to
any data integrity problem occurring with SAFE=NO. Kinda like:
http://i.imgur.com/8hoLlx5.jpg
>> -----Original Message-----
>> From: Dyck, Lionel B. (TRA)
>> Sent: Friday, June 17, 2016 9:02 AM
>>
>> I like the idea of the RFE but could I suggest broadening it to all
>> BSAM/QSAM since any
>> program that writes output could do the same thing. Put the test in
>> theaccess method to
>> prevent writing a sequential dataset on top of a partitioned dataset.
>>
Ditto for changing attributes of an existing nonempty data set, PS or PO.
Attempts to guard against this at the utility program cause other collateral
damage.
Long ago I allocated output of a utility (IIRC, GIMDTS) to a UNIX file with
PATHOPTS=OWRONLY because it was intended only for output in that
job step. Got "Permission denied". ??? PMR. WAD. Turns out the utility
always first opens its output data set for input so it can examine control
blocks to ensure that it is not DSORG=PO with no member name specified.
How many other places is this test performed incorrectly? If it were
done in a single place in the access method it could be coded
competently and not interfere with UNIX basefiles with PATHOPTS=OWRONLY.
(z/OS UNIX itself prevents sequential writes to directories. The same
should be done by access methods for legacy PDS. What happens if
one tries to overwrite the directory of a PDSE?)
BTW, I tried recently and I was able to read (IEBGENER IIRC) a UNIX file
allocated PATHOPTS=OWRONLY. It's easy to suspect that IBM disabled
a proper check because so many of their other products were doing the
wrong thing that it was impractical to fix them all.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
