On 26 April 2017 at 21:15, Paul Gilmartin <0000000433f07816-dmarc-
[email protected]> wrote:
> >Not explicitly. But I can't imagine a product the customer knows is going
> >to run as a z/OS started task with significant performance requirements is
> >going to decide to put a transaction-containing file on an NFS server on a
> >Linux box across the country. Nor, for that matter, are they going to make
> >it a temporary file.
>
[That didn't exactly come out unconfused, did it...I guess the gist of it
came across]
> SVC 99, I'd be less certain:
> o A socket can have a descriptor.
> o DYNALLOC can allocate to PATH('/dev/fd/'descriptor). I've done this
> with unnamed pipes created after step initiation; never tried with
> sockets.
>
It's not just the allocation, but the access methods. Are you going to open
your socket with QSAM? VSAM? EXCP? The $$$$ Media Manager?
>
> Are you suggesting that SVC 99 can allocate only objects that could as
> well have been allocated with JCL?
>
I don't think the list is much if at all different. Certainly there are
DYNALLOC *keys* that don't have a match in JCL, but those are mostly for
returning values. Are there different object types? I don't know.
(Things that can be allocated with JCL but not with DYNALLOC include passed
> data sets. Probably.)
There are some DYNALLOC keys missing for things that are inherently batch
JCLy; DD DDNAME= and of course DD * or DATA, and one or two others.
Tony H.
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
