At 02:13 PM 8/24/97 +0300, era eriksson wrote:
On Sat, 23 Aug 1997 18:02:02 -0400, Stan Ryckman
<stanr(_at_)sunspot(_dot_)tiac(_dot_)net> wrote:
I think whoever told you that was some sort of DOS-moron who thinks that
DOS/Windows defines "must be".
Explain why that would be true (lf+cr) since UNIX machines were the only
ones using internet mail for many many moon.
For the record, this is also not true.
Why, I don't know. Probably someone powerful with a brain-dead OS
insisted on it when the standards were being negotiated. It does
seem kind of silly to waste bandwidth on all those useless CR bytes,
but that's not going to change now. Imagine the mess.
RFC822 is dated 1982, at a time when I'd imagine TOPS-20 (CRLF??) was
a pretty big player. Not to mention MVS (? LRCF?? + 0x6f0f7f and stand
on your head??), Multics (LF??), even CP/M (CR), and later VMS (CRLF?).
RFC959, the ftp spec, is from 1985 and discusses implementation issues
on Multics and TOPS-20 as well as Unix and "IBM mainframes".
Now that I think about it, this must actually go back way before RFC822.
Somebody implemented the first ArpaNet that way, no doubt, and after
that it became a matter of pragmatics: "You want to talk to us,
then you speak our language." Even if done on UNIX systems, I'd
guess something like the following: developed using a serial port,
set to do LF<->CRLF conversion, so that a dumb terminal stuck on the
line could be used for monitoring and debugging.
I'd be interested if anyone has pointers to "history of ArpaNet" or
something like that. Facts are sometimes more interesting than
speculation.
Oh, this is the *procmail* list? In that case, I'll mention that no,
procmail recipes do not need to concern themselves with inserting
or removing the CR's associated with internet mail transport.
Cheers,
Stan.