Message-ID: <AlF=HcW00WBwA12HhU@andrew.cmu.edu>
Date: Mon, 11 Mar 1996 18:19:04 -0500 (EST)
From: John Gardiner Myers <jgm+@CMU.EDU>
To: drums@cs.utk.edu
Subject: Re: Free insertion of linear-white-space
In-Reply-To: <MailManager.826569321.18793.mrc@Ikkoku-Kan.Panda.COM>
References: <MailManager.826569321.18793.mrc@Ikkoku-Kan.Panda.COM>

Mark Crispin <MRC@Panda.COM> writes:
> > Existing
> > implementations don't generate whitespace between the field-name and
> > colon and many existing implementations can't handle whitespace
> > between the field-name and colon.
> 
> I've encountered such implementations.  I don't think you can say it never
> happens, just that it is rare.
> 
> I have no problem with banning this practice; however, it should be of the
> form "don't do it, but you should parse it if you see it".

I have an IMAP/POP server, non-trivial deployment, that bounces
messages if there is whitespace or comments between the field-name and
colon.  Similarly, munpack considers such lines invalid header syntax
and treats them as if they started the body.

To date, I have received zero problem reports about this.

Requiring all implementations to parse this kind of stuff has a
non-trivial cost in terms of complexity.  In some of these cases, the
cost far outweighs the benefit gained.  We are engineers--we have to
evaluate the relative costs and make appropriate trade-offs.

Keith Moore <moore@cs.utk.edu> writes:
> > > This is an (unmentioned) incompatible change in syntax which would
> > > outlaw  "this".is."legal".under."RFC822"
>
> I've certainly used this syntax before when mapping DECnet mail addresses to 
> RFC822.  I changed to use the percent-hack when I found that one or two SMTP 
> servers wouldn't accept it in a RCPT command.

Looks like you ran into the exact problem I was trying to fix.  The
above local-part, while legal in 822, is *not* legal in 821.

Requiring 822bis parsers to handle that syntax will necessarily
require a large number of them to *convert* such addresses into legal
821 syntax.  This has a relatively high cost.

-- 
_.John G. Myers		Internet: jgm+@CMU.EDU
			LoseNet:  ...!seismo!ihnp4!wiscvm.wisc.edu!give!up