On Wed, 3 Feb 2016 19:53:48 -0800, Skip Robinson <jo.skip.robin...@att.net> wrote:
>Sweet. Did not pick up on that. All the more reason to prefix symbols with a >unique string. > >> -----Original Message----- >> On Behalf Of Anthony Thompson >> Sent: Wednesday, February 3, 2016 07:48 PM >> >> Please note that with z/OS 2.2 the length of system symbols names has >> increased from 8 to 16, and may include the underscore character. >> >> Ant. >> >> -----Original Message----- >> On Behalf Of Paul Gilmartin >> Sent: Thursday, 4 February 2016 12:03 PM >> >> On 2016-02-03 19:27, Skip Robinson wrote: >> > This is why I strongly recommend that installation-defined symbols be >> prefixed with a unique string, which I also recommend be the SHARE >> installation code. It reduces the number of meaningful character to 5 or 6 >> but >> pretty much rules out stepping on toes. Debugging problems caused by >> symbol 'overlays' could be excruciating. >> > >> The namespace is too small in this 21st century. >> >> -- gil In the meantime, we still have the 8-character limitation. And whether we try to convert existing symbols to use a site prefix now or wait until we have 16 characters available, that's no small feat. Naming conventions of any kind, once they take hold, are time-consuming to change. What if the installation is not a SHARE member? We already use meaningful (to us) prefixes to make symbols intuitive to us without stepping on IBM symbols, let alone other vendors. Further, in the case I pointed out previously, the vendor decided to use symbol names that conflict with IBM filter criterion, and could conflict with IBM symbols. We have no control over such situations. Even if they don't conflict, should the installation be JES3, the changes won't get picked up, unless the product is issuing a *S <main>,CONNECT under the covers...more shenanigans. I understand that with the advent of SETLOAD IEASYM, this has changed a little, but at z/OS 1.13, the vendor clearly is not using this facility to effect the change. I stand by my earlier statement, no supplier should be dynmically inserting symbols into the table without express consent of the installation. In other words, there should be a configuration switch for which the default setting should be OFF. Better to document any symbols and have the customer define them explicitly at IPL via IEASYMxx. I have a counter-proposal. IBM can reserve certain prefixes or even complete names, as they do for SYSMOD IDs, that are off-limits, or use-at-your-own-risk. They can further set up a registry, as they do for module names, and while this registry is voluntary, we can certainly encourage vendors to participate. Regards, Art Gutowski General Motors, LLC ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
