todsn option idea.

[JohnMcKown]

This is perhaps "insane". But after a question and some responses on the TSO-REXX forum, I had a weird idea (again). How difficult would it be for the "todsn" command to be able to maintain ISPF member statistics. I don't know how the command is currently implemented. I do understand that the ISPF statistics are kept in the directory and that, basically, means that the STOW macro needs to be used. I would like something similar to what FTP does when QUOTE SITE ISPFSTATS is in effect.

If not todsn, what about your enhanced sftp? I think that it can at lease write MVS datasets. I'm not sure about reading them. Perhaps a
ls /+ispfstats=on to go with ls /+mode=text ??

The step which currently creates the PDS member via JCL // DD DSN=PDS(MEMBER) could simply write to a /tmp/some.file and then use todsn (or ftp or sftp) to transfer the data into the appropriate PDS member. Yes. a bother, but maybe of some interest.

--
John McKown

[dovetail]

John,

We are well aware of this requirement. Unfortunately, we currently use the IBM C-library for dataset I/O, and there is no facility for updating the directory when closing the member. (Pls submit your requirements to IBM, since this is a problem IMO).

So, assuming that we continue to use the IBM C library, we would have to reopen the dataset and update (rewrite) the directory. This may have performance / serialization issues. What is your opinion on the technical challenges?

[john.mckown]

Good question on rewriting the directory. I know that can be done with a STOW REPLACE type operation. But I don't do BPAM much.

I am actually in the process of abandoning PDSes in favor of z/UNIX files to the extent that I can. I've moved all my JCL, except for PROCs accessed via a // JCLLIB statement, and all my HLASM and COBOL source into UNIX files. I can still use ISPF to edit them. And grep to find stuff. I keep a z/UNIX shell up via ssh all the time now. I even use z/UNIX "vi" on rare occassion to edit files.

So, ISPF statistics are becoming irrelevant to me. Wish I had a "full screen" SDSF available under z/UNIX.

[dovetail]

For many years, we have placed our source (C/C++/ASM) in z/OS Unix filesystems and used "make" to compile/assemble/link. You can either link directly to a PDS or use "cp -X to install load modules in PDS(E)s.

We use Eclipse for nearly all of our editing, and then we use an Ant script under Eclipse to incrementally upload any changes to z/OS and to invoke a build via an ssh command. Its one click, and works quite well. Since all of our source and build/test scripts are in Eclipse, its easy to use CVS or SVN for source code control.

Of course, we still use a TN3270 emulator for ISPF and SDSF, and ssh for a z/OS Unix shell.