On Mon, Jul 13, 2009 at 9:43 AM, Kris Buelens<[email protected]> wrote:
> SYSPROF EXEC was a great enhancement.  Before there was only assembler code
> (DMSINS).  SYSPROF was the step IBM made to make the CMS startup
> tailorable.  Only later people started thinking: SYSPROF is IBM code one
> should not change.

It's a gray area and things evolve over time. When you start shipping
working sample code for an exit that a customer may fill in, the risk
exists that critical function is shipped in it and you require the
customer to rework his exit with an upgrade (or miss function).
Ideally the product should work without the exit filled in, and the
interface should not changed

In IGS the major modification to SYSPROF was to add a dozen hooks to
calls a local exit that was filled in. However, there's many fingers
in the pie and the interface is so delicate that the IGS platform
shipped with a standard local exit that called yet another local exit
that the installation could fill in. But again, that was sufficiently
complicated that they ship a boilerplate sample where you stick in
local processing again.

We had a lot of fun moving the customization into VMLINK - there's
neat things you can do with pre-exit and post-exit that are pretty
isolated. So the code invoked by SYSPROF and friends does a 'VMLINK
ABCTOOLS  ( FILE ABCDISKS' and that VMLINK control file described
which disks to access, at what file mode, including processing that
determined the right disk based on the user's credentials, picked up a
backup disk when the primary was not available, etc. The advantage is
that you don't have to predict all the possible intervention, but the
caller will simply "link the disk" and you make it as complex as you
need (and resist the temptation to write REXX variables into the
caller environment to make up for restrictions in the interface).

Rob

Reply via email to