This a copy of something I sent to the XMAILER list a few days
ago, but I think it's relevant to the discussion here too.
 
== Forwarded Mail ==
 
> Why invent a method of segmenting records which isn't fully
> transparent?
>
> Why invent yet another method of segmenting records since several
> already exist?
>
> (Go look at the varous MVS, VM, LISTSERV, and NJE formats).
>
> If you really wanted something standard you might come up with a
> way of sending "real ASCII" messages over BITNET (no line length
> etc).  This could be used for "Internet/RFC822" mail now and
> X.400 mail in the future (X.400 is also ASCII).
 
The NJE protocol used on BITNET has always allowed for a maximum
line length of up to about 32K, which should be adequate for most
mail messages.  The 80-character restriction comes from the Columbia
Mailer's use of 80-character punch files for mail.  This probably
arose out of RSCS Version 1's inability to handle punch files with
lines longer than 80 characters and print files with lines longer
than 132 characters (or whatever the line length is for the type of
printer that RSCS thinks the file will utilimately go to -- 132 is
the minimum).  I'm not an RSCS expert, but I believe that at least
some of the line length restrictions have been relaxed in RSCS
Version 2.  JES2 has always allowed 32K.  I'm not sure about the
various other products (JNET, etc.).
 
Of course, I realize that there are a lot of other complaints about
RSCS Version 2. :-)