On 2020-08-20, at 12:52:07, Alan Altmark wrote: > > On Thursday, 08/20/2020 at 04:43 GMT, Paul Gilmartin wrote: > > >> With what restriction on the compound symbol namespace? >> Of course if all supported tags are legal compound tails >> there is no restriction. > > The name space of REXX variables is a subset of tag names. You can put > almost any character, printable or not, in a tag name; not so much with > REXX variables. > Whereas the name space of compound tails is almost unlimited (save for an overall limit on the length of the resolved symbol name.) There's a considerable argument here for using compound symbols rather than simple symbols.
> In theory that's a problem. In reality, it isn't. I've been converting > tag names to REXX variables (tail or simple) for 30 years and it hasn't > tripped me up so far. > > However, given enough monkeys and typewriters, you get TCP/IP APAR PH06391 > (3/2019). It was a problem where someone put a comment in the DTCPARMS > files as :* instead of .* and TCPRUN tried to set a REXX variable name > with an asterisk in it. Oops. So now it ignores tags that aren't made > from characters REXX allows. User error, but we changed the program to > be more tolerant of human frailty. > Did they also start checking the return code from EXECCOMM? "User error"? It's pretty harsh when a user error "causes initialization of a TCP/IP virtual machine to fail." But the APAR doesn't state how the error was reported prior to the fix. I'd say the preferable fix for an unreported syntax error is to report the problem in a lucid message, then fail the initialization. -- gil