On Wed, 10 Jul 2024 13:44:49 -0400, Rick Troth wrote:
>This thread is relevant to IBM-MAIN because it discusses file handling,
>specifically in CMS.
>
To me, "mainframe" does not imply "MVS". If other contributors
wish it otherwise, they should petition for a charter change and
that hardware, Assembler, CMS, VSE, Linux for z, etc. be
deported to other suitable lists.
> ///
>WARNING:
>Logic which depends on invisible mark-up is dangerous.
>
IIRC, that was your habitual rant decades ago. I thought
you had got better lately.
>That blank after the dashes is easily stripped by any number of processors.
>
No! Leave my data alone.
>In the days when MIME was being cooked-up, I appealed to the email lords
>"please let white space be insignificant". They refused.
>So now, the divider between header and body *must* be CR/LF/CR/LF. In
>other words, it cannot be CR/LF/whitespace/CR/LF.
>("Insignificant" is not the right word. Linear white space is widely
>defined as one or more blanks and/or tabs. Let the reader understand.)
>
>This was particularly frustrating when I was a systems programmer
>running VM/CMS.
>CMS does not have the concept of an empty record. A record must have at
>least one byte. (Blank lines contain at least one space character.)
>This is arguably stupid,
>
Does BFS provide a solution? Otherwise, CMS should provide YA
filesystem which is tolerant of empty lines.
Long ago, I did an experiment, a single-byre zap to nucleus,
(changing a BZ to a NOP.) Then I was able to write and read
files with my assembler programs. Alas, many CMS utilities
program checked reading them. Is suspect that resulred
from subtracting one from l length, then EX MVC.
> ... but it's history (like tabs in makefiles), and
>
That was thoughtless design. Some implementations of make
accept 8 blanks as an alternative. There's probably other
consequential breakage.
>we had to take special steps in CMS land when processing email.
>Thankfully Pipeles *does* allow an empty record, but that's a whole
>nutha story.
>
CMS is the outlier here. I have imagined it could fall in step by
relying om NETDATA format or using explicit CRLF in its work
files so no empty records appear.
>For further reading, consider the message delimiter in traditional Unix
>mailbox files. (Not on-topic w/r/t lwsp, but another example of poorly
>thought precedent.)
>
That's generally solved by encoding lines that might otherwise
spoof that delimiter: "%46rom ..."
--
gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
