Date: Wed, 24 Jan 1996 14:05:30 -0800 (PST)
From: Mark Crispin <MRC@Panda.COM>
Sender: Mark Crispin <mrc@Ikkoku-Kan.Panda.COM>
Subject: Re: A suggestion for IPv6 domain literals
To: "Eric Norman (MACC)" <EJNORMAN@macc.wisc.edu>
Cc: drums@cs.utk.edu
In-Reply-To: <26012416035283@vms3.macc.wisc.edu>
Message-Id: <MailManager.822521130.18416.mrc@Ikkoku-Kan.Panda.COM>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII

On Wed, 24 Jan 96 16:03 CDT, Eric Norman (MACC) wrote:
> I agree that logging IP numbers is quite valuable; however, that has
> to do with the format of addresses in RFC822 or RFC821.

Uh, where do you think this is logged?  Don't you think that Received: headers
are part of RFC821/RFC822?

> When the DNS is broken, what problem does having domain literals solve?

Here's another example.  My machine asks the DNS for its own name, and gets an
error.  So, it sets the From address of my mail to be "mrc@[192.107.14.129]"
which is perfectly valid *and* replyable.

Otherwise, the email application has to say: "Horrible error 69: I can't get
my host name.  Go away."  I don't think that is an improvement.

> And for whom?  Can't such problems be solved another way?

One solution would be to put Eric Norman on 24 hour international beeper call
with the task of making sure that all DNS servers worldwide are up and
functioning, and take the heat when they aren't.

The point is that we have a mechanism that works; it is appropriate for the
problem, it scales, and it is at times a lifesaver.

> I'm not
> claiming that I know the answers, but I think such questions need to
> be asked and addressed.

These questions have been asked and answered repeatedly for at least the past
20 or so years.  This doesn't mean that novices who ask it should be given the
brush-off.  However, novices should recognize that they are novices; and ask
questions, not go off and make dangerous proposals.

> My motivation for suggesting elimination of some things is to try to
> keep things as simple as possible.  I believe that making things simple
> will have a significant contribution to making the electronic mail
> system easy to understand, reliable, interoperable, have more "working"
> inplementations, and all that good stuff.

This is a handwave.  Simple for who?  The user, or the implementor?  If you
want "simple for the user", then you want a system which has robustness and
does not collapse due to a failure of some name server.

I have no sympathy for those who want "simple for the implementor".  Internet
email is simple enough that any competant programmer can produce a functioning
email implementation.  I don't really want to make email simple for
incompetant programmers.

> By "simple as possible", I *do not* mean reducing thing to the lowest
> common denominator.  What I mean in this case is that things that are
> rarely used and only in exceptional circumstances should not be part
> of the standard if there are other ways to solve the problem.

But, as we have tried to tell you, domain literals and group syntax are
commonly used.  *You* may not use them, but you can not generalize your
personal anecdotal experience to anything else.

Your personal anecdotal evidence may be that your mailer's Date headers:
	Date: Wed, 24 Jan 96 16:03 CDT
work alright, but in less than four yeare they will not.  Two-digit years and
the use of alphabetic CONUS timezones such as "CDT" (are you really in summer
time during the middle of winter?) were deprecated years ago.

Suppose all your email started bouncing with:
	Horrible error 69: two digit year "96" invalid
	Horrible error 105: timezone "CDT" invalid
You'd be awfully upset.  Now, this is something that's clearly broken, and a
superior replacement exists.

There are *no* replacements for domain literals or for group syntax.  You'd
have to invent a new replacement.  And there would have to be a transition
period of a decade or more during which both are permitted.  I don't call that
a simplification.

Domain literals and group syntax work, and work well.  I see no reason to
change either.

> > I object *very* strongly to any move to deprecate group syntax.
> I suggested elimination, not deprecation.  Since people are using it,
> I'll suggest deprecation instead; I thing mailing lists are a better
> solution.

No, mailing lists do not solve the problem.  Group lists give you the option
of an empty membership and *non-replyability*.

This is an old capability that has been around for over 20 years.  There is no
excuse for any software that does not support it.  Group syntax has a definite
and unique use, and I will fight tooth and nail to stop any proposal to
deprecate it.