Message-Id: <200003171449.JAA22141@astro.cs.utk.edu>
From: Keith Moore <moore@cs.utk.edu>
To: Charles Lindsey <chl@clw.cs.man.ac.uk>
cc: DRUMS <drums@cs.utk.edu>
Subject: Re: Is it good to publish a new RFC on SMTP without internationalisation? 
In-reply-to: Your message of "Fri, 17 Mar 2000 10:31:05 GMT."
             <200003171031.KAA07932@clw.cs.man.ac.uk> 
Date: Fri, 17 Mar 2000 09:49:37 -0500
Sender: moore@cs.utk.edu

> > no it wouldn't.  because it implies having two forms of each email 
> > address (one ascii and one utf-8) and some sort of conversion at 
> > the boundaries, with the usual bitrot due to implementation errors.
> > and when addresses leak  across the boundaries (which they will)
> > chances are they won't be usable from the other side. 
> 
> Eh? Why would that be. Ascii is a subset of UTF-8. If you look closely,
> you will see that all the headers of this message are already written in
> UTF-8.

yes, but if the headers of your message had used characters with
values > 127, my mail user agent would have (correctly) refused 
to generate a reply to your message.

> So if people want to move to domain names or local-parts using exotic
> characters, then UTF-8 will, I am sure, be the way we eventually go.

you might be sure, but to me it's clear that using utf-8 on the wire 
for email (at least, any time in the next ~20 years) would be an 
incredibly stupid and disruptive thing to do.  so I hope you're wrong.

> But if someone wanted such things to pass through present transports,
> then said transports might try to squeeze them into RFC 2047 format,
> which might or might not work, but would sure lead to an unholy mess.

yes.  2047 is another naive approach that is often suggested.
it would indeed lead to an unholy mess.  

given the transition issues, some sort of encoding scheme is a given.
but for a variety of reasons it cannot be 2047.

> So I hope no-one will try to do such things until transports are 8-bit
> clean as a matter of course. That sure is not going to happen overnight,
> but it seems we ought to be preparing the ground by starting to move in
> that direction, even though it might be 10 years before we can reap the
> benefit.

let me get this straight: we don't know where we're going, but we 
ought to start moving in that direction anyway?

> FWIW, Netnews is already planning to go that way.

Netnews != email, and despite superficial similarities, what makes sense
for netnews doesn't necessarily make sense for email.

Keith