On Sat, 25 Mar 2017 13:35:52 -0500, Bill Woodger wrote:
>
>I like pipes. I don't like that little caveat you made, except why would it be
>relevant here?
>
It's probably not relevant, and you seem to understand. But others may need
more.
When the world was young and DDNAME support by FTP was newly announced,
I put on my Black Team cloak and tried something like:
//EXEC PGM=FTP,PARM=SOMEWHERE
//SYSUT1 DD PATHOPTS=ORDONLY,PATH='/etc/services',
// RECFM=FB,LRECL=80, (Is that big enough?)
// FILEDATA=TEXT
//INPUT DD *
binary
put dd:SYSUT1
quit
Expecting that FTP client would OPEN a DCB on SYSUT1 and do QSAM GETs.
FILEDATA=TEXT would instruct the access method to separate records at
<NL> characters. RECFM=FB,LRECL=80 would instruct the access metnod
to pad each record with x'40'. FTP client should write that data to a socket
unmodified, leaving the padding blanks and adding no <CR><LF>.
This should be similar to the behavior of PGM=IEBGENER (mutatis mutandis).
Instead, FTP discovered somehow (TIOT? JFCB?) that SYSUT1 was a UNIX
path; ignored most of the DD statement, and transferred that directly, binary,
<NL>s and all.
Grrr.
It might also fail if SYSUT1 referred to a pipe because FTP supported only
regular files. But more recently, FTP announced support for named pipes.
If that works, one might even submit a batch COBOL job to write to the
input end of a named pipe and fetch the data to desktop interactively
with Filezilla or the like.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
