On Wed, 22 Jan 2014 07:48:42 -0800, Skip Robinson wrote:
>Good catch. In my recent SHARE pitch on system symbols, I strongly
>recommend that all installation-defined symbols be a full eight characters
>long regardless of initially anticipated value. In addition, I recommend
>that all such symbols be prefixed with an installation identifier, such
>as SHARE company code, to clearly identify them as user defined. This
>practice will also group installation symbols together in a D SYMBOLS
>display.
>
Where are the SHARE company codes listed? What is their domain of
applicability? Do these exist in parallel and in contention with IBM
registered component prefixes.
In a universe with less archaic length restrictions, the custom is to
incorporate
a registered domain name, rewritten big-endian. IBM shows some slight
adherence to this in such as:
/usr/lpp/booksrv/cgi-bin/com.ibm.bkmgr.CgiJavaBridge.jar
/usr/lpp/smp/classes/com/ibm/smp/GIMJVCLT.class
What!? not "com.IBM"?
I surely wish SMP/E SYSMOD IDs were so flexible that we could
incorporate a corporate ID (preferably domain name) in PTF
names.
>From: Tom Marchant
>Date: 01/22/2014 05:42 AM
>
>>SYMDEF(&IP1='121.122')
>
>No. As documented, "The length of the resolved substitution text cannot
>exceed the length of &symbol, including the ampersand on &symbol and
>excluding the single quotation marks on 'sub-text'."
>
>So your example is not valid.
>
An abomination; it ain't that hard to code a routine that substitutes
values longer than their names. Doesn't that work for JCL symbols?
But they can *never*fix*it* if, as I assume, the symbol substituting
facility has no way to report a buffer overflow, and (some) callers are
in no position to handle an error if one were reported.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
