David Crayford writes:
>Do you have any "real" world experience with open source and
>porting to z/OS?

Yes, some.

Tony Harminc opines:
>Sure, if all your application does is crunch numbers or manipulate
>bytes. But if it has any interaction with the operating system such as
>calling its services...

Which (other) services?

>referencing external names (dsnames and file names being only two of
>many),

There is no requirement on z/OS to support dataset names unless operating
on datasets (VSAM files, PDSes, PDSEs, etc.) Which I've already covered in
my statement.

>dealing with operator services, etc. etc. etc., you will find it

There is no requirement on z/OS for open source ported applications to
"deal with" operator services. For example, there is no requirement on z/OS
that applications produce SMF records.

*Requirement*. Everything you're describing may be desirable, even highly
desirable, but not REQUIRED.

Now, I stipulate that there are many desirable capabilities. Operating
on/with EBCDIC data is often useful. There are two ways to try to
accomplish that goal:

1. Complain to and criticize each and every individual open source
community for every open source product, that they must accept adding a
laundry list of features to their mainline source code in order to exploit
z/OS-unique and z/OS-desirable features. That approach doesn't seem to be
working, does it?

2. As the Java community has, and IBM has (and continues to do), bolster
the generalized runtime environments on z/OS so that more open source
products can come to z/OS more easily and (optionally!) exploit z/OS
without changes to the mainline source code (or with at least fewer
changes). For example, if you want open source applications to be able to
operate on/with EBCDIC data, how about a generic approach that works for
all (or at least most) open source products? My /ebcdic path idea isn't
necessarily best or even viable, but at least I'm trying.

Might I suggest door #2? What would a "Common Transparent z/OS Services
Environment" look like? I'm not excluding the possibility that IBM might
implement *some* of that sort of stuff. IBM already is, e.g. z/OS 2.1's
GNUish file I/O.

For example, one radical (but probably viable) idea would be a userland
GNU/Linux atop z/OS. Anybody interested in doing that?

--------------------------------------------------------------------------------------------------------
Timothy Sipples
GMU VCT Architect Executive (Based in Singapore)
E-Mail: [email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to