On Fri, 29 Jan 2016 08:52:00 +1100, Peter Morrison wrote:
>
>I have been following the COBOL V5 debate with interest,
>
>Currently, when using the binder to create a program object, the output load
>library MUST be a PDSE.
> 
No.  (I believe)  It might instead be a UNIX directory.

>There is a lot of animosity about PDSEs.  They have been around for many
>years.  But, when it comes to sharing them across systems and/or Sysplexes,
>IBM still haven't gotten it right.
>
UNIX directories can be shared across systems with zFS sharing and across
plexes with NFS.  Even to non-z systems.  IBM got that right; integrity is
maintained.

>PDS's can be put on a shared pack and be shared across multiple systems
>and/or Sysplexes.  Of course, you had better not try to update them
>simultaneously from several places or bad things can happen. Sites that do
>the sharing usually have procedures in place for that.
>
>Looking at the IHAPDS map (in SYS1.AMODGEN), which maps a load module (or
>program object.) PDS entry returned by BLDL, and the directory data on disk,
>there is a flag that can be used by a program to distinguish a program
>object from a load module.  There is an optional extension that has larger
>offsets (greater than 3 bytes.)
>
>SO, what if that bit was set in a directory entry on disk in a PDS load
>library?  That could mean that the actual data in the PDS member was a
>program object, not a load module, and be handled appropriately by the
>system loader.  Then, you would get all the advantages of a program object
>with none of the problems associated with a PDSE.  (The issues associated
>with long alias names are not covered here.  But note that when it comes to
>execution, you need an 8-character name anyway - look at the LOAD etc.
>macro(s) for example)
>
>(The binder could have a special option that would tell it to put a program
>object in a PDS - to override the default of putting a load module there)
>
The bad news:  UNIX directories must not appear in STEPLIB (I think;
I haven't tried it) nor in LINKLIST.  They oughta fix that.

-- gil

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

Reply via email to