[Top] [All Lists]

RE: Non-ASCII Internet addresses?

1993-05-03 16:00:34
This proposal will do nothing to increase the ability of people to
communicate over the Internet.  In fact, it will make it more difficult for
them to do so.

This point has been brought home to me rather forcefully in recent weeks by the
troubles we've had in getting support for the existing header extensions as
well as the SMTP service extensions into place. Despite years of experience in
this area, I among others seem to have underestimated the difficulty by at
least an order of magnitude.


Everyone in the Internet will suffer if such a proposal is adopted.  What's
worse is that no new functionality is gained for that pain.  A significant
fraction of the Internet may 'benefit' because they can use their login id
for their email address.  However, the majority of the Internet will gain
nothing.  And everyone will lose because there will be another piece of
"black magic" that users have to know to use email effectively.

Making e-mail addresses more complex (even if this translates to simply making
them longer) is never a good idea unless there's a huge gain at the end of the
effort. Such a gain was in fact realized the last time the structure of email
addresses was seriously tampered with (i.e. the addition of domains).

And as the experience with domains shows, such an extension is sure to run up
against limits in something somewhere, and all we get out of it is another
exercise in finger-pointing and eventually another set of hacks to work around
the problem area. Words simply cannot express my distaste for such a result.

The beauty of the existing header extensions is fourfold: (1) Users don't have
to do anything different to send mail, (2) There's no mandatory impact on UAs
or MTAs, (3) Infrastructure is only stressed in terms of overall header
lengths, which while a problem area is known to be considerably more tolerant
that most other things (address size in particular), and (4) Even if it does
break our too-fragile infrastructure we only lose headers, which are not
strictly essential.

None of these apply to extensions to the mailbox parts of addresses.

And despite all this, getting these extensions developed and deployed is
proving to be a monumental task.

By all means, let's work on a directory service that allows us to address
people by their real names, and let's interface it to our UAs.  When that
happens there will be some pain, but we will gain something incredibly
useful in return.

I agree completely with Keith on this point.

More specifically, while continuing this discussion is fine and dandy, it
should be understood that none of this should be taken as contributing in any
way to the existing MIME documents. If this group goes ahead and proposes an
extended format for mailbox specification (which it is certainly capable of
doing despite Keith's and my position) it will have to be done as a separate
document starting out as (at best) a proposed standard.

If we do decide to extend mailbox formats, let's at least wait until there's
some minimal deployment of the existing extensions we've done, and we can more
adequately assess the impact on the infrastructure of such changes.


<Prev in Thread] Current Thread [Next in Thread>