On 2017-12-20, at 09:34:06, John McKown wrote:
>> 
>> Only with BPXWDYN can I allocate a mixed concatenation of Classic
>> data sets and UNIX files/directories.  COBOL ASSIGN?
> 
> ​Nope - I just double checked the documentation previously posted. The
> COBOL dynamic allocation via the ASSIGN can be _one_ DSN xor _one_ PATH​,
> not a concatenation.
>  
I bet bad things would happen if one tried multiple ASSIGNs and one
BPXWDYN CONCAT.  And BPXWDYN RTDDN is precious for ALLOCing additional
catenands.  Or avoiding collisions.  I shudder whenever I see a
programmer using RANDOM to generate a ddname.


On 2017-12-20, at 04:13:02, John McKown wrote:
> 
> ​Which "production" languages would that be? Enterprise COBOL and PL/I
> _both_ support dynamic allocation of files.
>     ...
> ​HLASM doesn't. REXX does & doesn't (EXECIO won't but you can use BPXWDYN
> before EXECIO). I can't find the z/OS FORTRAN documentation right now.​
>  
Aah!  You mean built in to the read/write.  40 years ago, FORTRAN had
no OPEN.  It simply assumed that the competent operator had mounted the
input tape om physical unit 5 and the output tape on physical unit 6.
Additional tape addresses ad lib.

> Most of the z/OS UNIX language will because they use the C "fopen()".
> Interestingly, the Rocket Software Python port does not at this time. But
> then, it doesn't support DD statements either, only UNIX files (I guess via
> C's "open()" instead of "fopen()").
>  
It's a shame that z/OS doesn't (simulated?) mount the entire ensemble of
catalogued data sets at a mountpoint above root so open() would just work.
Kinda NFS via loopback.

-- gil

Reply via email to