On Sun, 23 Oct 2016 15:21:33 -0500, Walt Farrell wrote: >> >>It's the curse of Conway's law. z/OS provides too many dissimilar ways >>of performing similar functions. Name--token, system symbols, >>jcl symbols, ISPF variables, ISPF Edit variables, Rexx variables, ... >>All have different syntax to set and access them. > >The situation is perhaps further complicated by multi-address-space >considerations, especially when invoking UNIX commands. > Perhaps I showed a bias in omitting mention of UNIX environment variables. *But* for those the rules are gorgeously simple. Variables not exported are private to the shell. Exported variables are inherited by all child address- spaces, never by sibling or parent address-spaces. The latter can be a nuisance, but that's what sockets and command substitution are for.
And, in POSIX shell at least, there are no "put" and "get" function calls; "=" and reference-to-name serve the purpose just as for any other symbols. (I very recently stumbled over the (needless, IMO) distinction between ISPF variables and Edit variables.) -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
