From: "Kari E. Hurtta" <Kari.Hurtta@dionysos.fmi.fi>
Message-Id: <199512020944.LAA11454@dionysos.fmi.fi>
Subject: Re: On the use of 8-bit transfer
To: presnick@qualcomm.com (Pete Resnick)
Date: Sat, 2 Dec 1995 11:44:33 +0200 (EET)
Cc: moore@cs.utk.edu, MAILNEWS-L@segate.sunet.se, drums@cs.utk.edu,
        ratause@dsv.su.se
Reply-To: Kari.Hurtta@dionysos.fmi.fi (Kari E. Hurtta), presnick@qualcomm.com,
        moore@cs.utk.edu
In-Reply-To: <v03003e00ace4e36030c3@presnick.qualcomm.com> from "Pete Resnick" at Dec 1, 95 09:02:46 am
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

[ I suppose that this is not longer in topic of these mailing lists.
  Notice Reply-To: -header. ]

Pete Resnick kirjoittaa:
> On 11/30/95 at 11:40 AM, Keith Moore wrote:
> >> If 8BITMIME extension is available, it will take care of conversion
> >> to 7-bit in event when some MTA does not support 8BITMIME.
> >
> >Not necessarily.  A server supporting 8BITMIME is not required
> >to perform such a conversion, though it will often be performed
> >in practice.
> 
> Ask a silly question and you get a silly answer, but:
> 
> If it will "often be performed in practice" (and it would probably be
> important to know how "often" that is), why can it not simply be required?
> That is, since we are *supposed* to be documenting practices on the
> Internet anyway, why not just make this a requirement of using 8BITMIME?

There is also bunch of cases where only possibility is bounce of message.
In these cases conversion is not possible (all possible results of conversion
are protocol violations (one protocol or another)).

I give following example in private mail:

{ Following text is answer to one question -- I deleted that question 
  because is is not now relevant. I'm talking here about situation where
  next host does not support 8BITMIME. }

Sendmail strips it to 7bit format (but also if sendmail convert it to
7bit format, it will produce protocol violation).

Pass following mail { Non-relevant headers deleted } with using
BODY=8BITMIME:

	MIME-Version: 1.0
	Content-Type: Message/x-my-bilatery-type
	Content-Transfer-encoding: 8bit

	{ 8-bit data in here }

Because quoted-printable and base64 are forbidded in subtypes of Message,
Sendmail can't do converion 8bit -> quoted-printable or 8bit -> base64.
Only correct action is bounce the message, but sendmail detect that problem
in too late (it have already writing to body of mail to next host). Result
is that that bodypart is _stripped_ to 7bit.

Notice that Content-Transfer-encoding: 7bit is only required for
Content-Types Message/partial and Message/external-body. So that example
is legal input.

Another case (what is possible to fix by changing of sendmail.cf) is:

        MIME-Version: 1.0
        Content-Type: Message/news
	Content-Transfer-encoding: 8bit

	MIME-Version: 1.0
	Content-Type: Text/Plain; charset=ISO-8859-1		
	Content-Transfer-encoding: 8bit

	{ 8-bit data in here }

That produce same result. (Notice that content-type Message/news is in
Assigned Numbers, so at least that content-type does not cause that this
is illegal input. That example may be illegal, bacause use of MIME in
news (message), but this is also very questionable). That case is possible
to fix by adding line

Cs news

to sendmail.cf. That tells to sendmail that Message/news should be treated
as Message/rfc822.

{ Last example and quotation from "Sendmail Installation and Operation Guide"
  deleted. }