> > is me (michael wagner)
>   is phil howard
    is me again
#   is Dennis Ferguson
 
  ( a discussion of a new continuation convention for 822 headers )
 
> >Whatever for? There already is a continuation convention
> Oh?  Surprise me... what is it?
>
> The only one I know about is inserting at least one blank on the
> continued lines,
 
  That's the one and only.
 
> which neither WISCNET nor FAL abide by under all circumstances for
> all headers.
 
  Don't get confused here.  The fact that some product or other
  doesn't implement a facility properly has nothing to do with
  whether the facility is properly and adequately defined.  When a
  product doesn't work, fix it (or, in this case, get it fixed).
  Why invent a new feature when an existing one is adequate?
 
> >   Starting in column 1 is illegal 822.
>
> Right, but tell that to WISCNET and FAL.
 
  Are we supposed to change lots of other software because neither
  Wisconsin nor IBM can read and understand a standard?  When
  vendors build things that don't work, don't use them.  Certainly
  don't propose switching an important network over to relying on
  them (back to my comments at the head of the article).
 
> > > Q: does RFC822 allow a line to be broken in the middle of a quoted
> > > field?
> >   Yes, of course. Quoting 822, section 3.4.5
> ...
> > > MAILER barfs on it.
> >   Why doesn't this surprise me?  (hmmm...wonder what UCLA MAIL does
> >   with it) (says he with a gleam in his eye).
 
  I tried this, by the way.  UCLA MAIL gets it wrong too.
 
> I'm sure we'd welcome a rewrite of MAILER.  Are you willing?
 
  Yes, of course.  I'd love to do it.  I'm sure I could do a good
  job, too.  Are you willing to pay?
 
  Before you jump to the conclusion that I am being conceited here,
  let me point out that I personally know several other people on
  the network who could do as good a job or better of rewriting
  MAILER as I could (and surely there must be some that I don't
  know :-) ).  One of the best candidates is the original author
  (Crosswell).  However, no one with control over the resources
  necessary seems to think it's important enough to provide the
  resources to get it done.
 
  In order to explain what I mean, I will quote from something that
  I sent as private mail in reply to Dennis Ferguson's posting a few
  days ago.
 
# The proper, standard solution is ..(description of solution)
 
  This is, of course, so obvious that one wonders why anyone even
  spends time discussing it.  It's a bug, and it should be fixed.
  However, because of the voluntary and anarchistic nature of
  BITNET, issues of only technical relevance are often resolved on
  other, less relevant grounds.  See discussion later.
 
# LISTSERV, unfortunately, does not follow this standard practice.
# Instead it ... (kludges described) (though, to be fair, I also
# understand that this is forced on LISTSERV by another bag of
# kludges, the Crosswell mailer).
 
  All, sadly, true.  The history weighs in here a bit.  In all
  fairness, Alan never wrote the mailer to be as widely used as it
  is, and it had a number of short cuts wired into it.  His
  installation seldom gives him enough time to work on it, and it
  hasn't kept up.  LISTSERV is similarly the voluntary work of one
  guy at one installation.  It meets a need, and so people run it.
  Wonderful neither of them are.
 
  What is really needed here is for someone to commit resources to a
  significant piece of software development and maintenance.  Those
  people who have a sufficient understanding of the problems to
  write the software are generally busy people who don't have the
  time to volunteer to do it out of good will (there are several
  man-years of work hiding out there in the general areas of
  replacing MAILER, LISTSERV and the various user agents with real
  programs, smooth migration, and maintenance).  And their
  installations generally don't have the money.  The organization of
  BITNET/NetNorth/EARN seems to be too loose to come up with the
  money and infrastructure necessary to house and retain people with
  sufficient expertise to do the job.  Non the less, I think the
  networks are going to have to find the resources to sponsor the
  work, or else flounder.
 
# Thus, expecting every UUCP site ... to modify their rejection
# notices to suit LISTSERV ... is more than a little parochial.  The
# real solutions, in order of preference, are (1) fix the damn
# mailer, (2) fix the damn mailer, (3) fix the damn mailer, or (4)
# add yet another kludge to LISTSERV to detect this week's problem
# rejection message.  Want to bet which fix is applied, if any?
 
  You are correct, of course, about the parochial-ness of the whole
  thing.  The answer to your question is clear: 4.  Eric is a
  volunteer who no one pays for his work, and he is therefore free
  to do what he wants.  No one seems to be capable of making
  technical decisions for the good of the network as a whole; his
  interests of improving his code come closest to be in concert with
  the network needs, and therefore kludging LISTSERV is the most
  likely fix.
 
  I don't think simply patching MAILER is feasible.  There are so
  many things that need to be done that a (perhaps only partial)
  rewrite is needed.
 
# I have a prediction, that by 1990 the only machines running
# LISTSERV will be otherwise unused 3090's since the reject filter
# will be so large that only these machines will have enough CPU to
# move the mail.  And, at the same time, every VM site will *still*
# be using the same broken version of the Crosswell mailer that
# they're running now (if they even understand BSMTP by then), and
# LISTSERV will still be sending stuff out with the wrong envelope
# address.
 
  I predict that those people who don't wake up and see the benefit
  of more standardized mail might well find themselves running post
  offices that implement such quasi-artificial-intelligence
  algorithms.  But it needn't be that way.
 
# Bizarre.
 
  Indeed.  But I'm not convinced that understanding BSMTP (821/822)
  is really any step forward.  There are as many implementations of
  that as there are versions of LISTSERV and MAILER in production.
  They all implement the same 'standard' differently.  Wonderful!
 
  I want to put in a plug here for X400.  Not the worlds best
  standard, but I think it is more in the right direction than
  821/822/920 and friends.
 
# Dennis Ferguson University of Toronto
 
Michael Wagner, University of Toronto (Bonn detachment :-) )