In a recent note, Ed Gould said:
> Date: Tue, 23 Jan 2007 07:50:04 -0600
>
> I believe there has been an option in SMS (since the early 1990's
> anyway) to write an eof into empty datasets, just by specifying it in
> the IGDSMSxx member in parmlib. IIRC this refers to any dataset that
> has not been openned that is SMS managed.
>
What I see is:
#<<3.2.6.3 "z/OS V1R7.0 DFSMS Using Data Sets"
______________________________________________________________
3.2.6.3 Factors to Consider When Allocating Direct Access Data Sets
When the system allocates a new SMS data set with DSORG=PS or no
DSORG, the access methods treat the data set as being null, that
is, having no data. A program can safely read the data set before
data has been written in it. This means the first GET or first
CHECK for a READ causes the EODAD routine to be called.
Hmmm. I had understood that this works only if DSORG is known.
But the wording appears back to the oldest OS/390 doc I find on
publibz. And note that the mechanism is not described. It could
be by writing an EOF at the beginning; it could be that a flag is
set (in the DSCB?) to indicate that the data set is uninitialized.
Note that it mentions GET and READ, but not EXCP. And it does not
mention any PARMLIB option.
For data sets other than system managed with DSORG=PS or null, the
program will receive unpredictable results such as reading residual
data from a prior user, getting an I/O error, or getting an ABEND.
Reading residual data can cause your program to appear to run
correctly, but you can get unexpected output from the residual
data. You can use one of the following methods to make the data set
appear null:
1. At allocation time, specify a primary allocation value of zero;
such as SPACE=(TRK,(0,10)) or SPACE=(CYL,(0,50)). This
I've done this; it works.
technique does not work with a VIO data set because creation
includes the secondary space amount.
Alas. It mostly works, but there's likely a performance impact.
In fact, I've run the system out of paging storage by doing this.
VIO ought to be more faithful in its emulation of real DASD.
-- gil
--
StorageTek
INFORMATION made POWERFUL
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
