On Fri, 13 Sep 2013 11:44:56 -0400, Charles Mills wrote:

>I have to say I agree. If PDSEs are fundamental to the OS, then internal OS
>code should support them.

It does.  But not early in the IPL process.  In fact, there are important 
system 
components that are designed to use facilities that are available only in 
program 
objects, not in load modules.  Those system components cannot reside in a PDS.

>If they're not fundamental, then COBOL should not be requiring them.

That is an excuse for you to be hesitant to use PDSE for your applications. 
Not a very good one, IMO.  Do you have a business case that requires that 
PDSE support be available earlier in the IPL process?

>-----Original Message-----
>From: IBM Mainframe Discussion List [mailto:[email protected]] On
>Behalf Of Clark Morris
>Sent: Thursday, September 12, 2013 6:29 PM
>
>
>>While I have been away from systems programming for over a decade, I am
>>appalled to see that a PDSE still cannot contain SYS1.LINKLIB, SYS1.LPALIB
>>and SYS1.NUCLEUS.

Do you have a need for it, or is that just some religious doctrine?

>>Would anyone care if NIP took a cylinder, 2 cylinders or even
>>100 cylinders.

Do you mean the IPL text?  I would care if the IPL text was 100 cylinders. 
That would be nearly 5 MB of code, and I would expect that it would need 
to have PTFs applied to it far too often.  Changing the IPL text is somewhat 
more complicated than changing elements in LINKLIB, LPALIB and NUCLEUS. 
In particular, SMP/E is not designed to replace the IPL text.  You can't look 
at your target zone in SMP/E and determine that NIP is at the correct level. 
Indeed, it is not easy to determine that the IPL text is correct.

If, OTOH, you are talking about the NIP modules that reside in SYS1.NUCLEUS, 
I guess you would still be "appalled" that SYS1,NUCLEUS still would not be 
able to be a PDSE.  But why do you care?  Do you have code that needs to 
be in the nucleus?

z/OS is a big and complicated operating system.  Like most operating systems 
today, it is implemented in several layers.  NIP has the most basic support 
needed to load the nucleus.

>>If started task is a good idea for PDSE, it should be equally good for VSAM.

Started tasks are required for many important system functions. 
It is not clear to me that the use of PDSE is one of them.

You can't do much with VSAM or most other data sets until the CATALOG 
address space is active.  Other started tasks are required for you to access 
SPOOL.  Or issue an ENQ.  Or allocate a data set for your address space.  
z/OS is not fully active and ready to process applications until after many 
started tasks are active.

>>At least code to read PDSE's should be available to NIP and maybe
>>SYS1.NUCLEUS.

Why?  Do you have a need for it?  What is your business case?  There are 
a lot of important system functions that are not available in the nucleus.

-- 
Tom Marchant

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

Reply via email to