Re: Problems with the 8-bit cleanliness document

[jwagner@xxxxxxxxxxxxx]

> From: jwagner@princeton.edu:
> 
> >  1.  Require all 8 bit mail to be originated in MIME format (my preferred
> >      solution)
>  
> I agree that this is preferred...
> 
> >  2.  Specify that 8 bit mail not in MIME format must be ISO-8859-1
(parallel
> >      to the specification for USASCII for 822).
> 
> This makes some sensem in that it is the only allowed unlabeled 8bit
> body, but then, does it not also have to be MIME labeled as being
> 8bit, vs 7bit, with a MIME label.  So how then is it not MIME?
> 
> What am I missing here?			  Cheers...\Stef

As the spec is currently written I can validly send headers in EBCDIC (as it
is just an 8bit character set).  Now this is not going to be very usable, but
it is legal.  It won't matter if the mail is in "proper" MIME format since no
other sites can decipher the EBCDIC headers.

Under 822 I know what the header character set is, under 8bit I don't. 
Requiring MIME format is a way to force the headers to be in a particular
character set.  If MIME format is not required for 8bit mail, then the
character set *for the headers* must be.

The basic problem is that there are two parts to successful mail transfer. 
There is the pipeline problem (do I have an 8bit pipe from source to
destination).  This is what gets addressed by the 8bit cleanliness proposal.

The second part of the problem is the "paper transfer" problem.   Can you
read what I wrote?  If I send you arabic and you can't read arabic it doesn't
matter that our fax machines could transfer the text.  A mail system that
transfers lots of 8 bit data with a general loss in readability is a failure
to the community.  This part of the problem is not addressed by the 8bit
proposal but was addressed in the WG draft.

I used EBCDIC in my example because it is an obvious example of an
incompatability.  But the blatant incompatabilities aren't the ones that will
drive us nuts.  It will the the small ones.  My system uses ISO 8859-1 and
yours uses 8859-2.  Although most character codings match, some don't and the
result may produce mail that doesn't interoperate.  The users get frustrated
because what they see looks almost right but the UAs and MTAs don't work.

It seems to me that we are performing a disservice to the network if we claim
SMTP supports 8 bit mail but we don't address the end-user to end-user
problem.  

   John