Re: Last Call: On-Demand Mail Relay (ODMR) to Proposed Standard

[Ned Freed <Ned.Freed@xxxxxxxxxxxx>]

> > this is an ancient discussion.  careful reading will tell you that
> > 'hostname' is not defind beyond the file hosts.txt which became obsolete
> > some years back.

> If this is an ancient discussion, I don't believe it is adequately documented.

I know of no formal specification that overrides the applicability of RFC952,
however silly it may be, to domains used in email. If someone is aware of
something of this sort, please point it out.

Note that the age of the document is completely irrelevant. The RFC index says
RFC 952 hasn't been obsoleted by anything, so it is still applicable when cited
in other specifications, especially full standards.

> Based on the citations I quoted, I believe I am quite entitled to draw the
> conclusion that I did, since material from RFC952 is used normatively by
> RFC1123 in a general statement about "syntax of a legal Internet host
> name", without any indication that this applies only to names in a
> "hosts.txt" file.  The definition in RFC952 is stated as an a priori
> assumption rather than a particular feature of the "hosts.txt" file.

And what follows from this is that on the operational front, there are plenty
of implementations out there done by people who have read the specifications
the same way you have and don't allow things like underscores in domain names
in email. And given that this is a perfectly reasonable reading of what's
there, you cannot claim such implementations are incompliant, much as you
(and I for that matter) would like to.

> Also, both RFC821 and <draft-ietf-drums-smtpupd-10.txt> define the syntax
> of a mail domain that is of the more restricted form.

Absolutely true as well.

Randy, I have no problem in principle with the idea of expanding the range of
legal characters in domain names used in email. However, if you want to do
this, you need to write a specification that does it and get it onto the
standards track. RFC2181 doesn't even come close to doing this.

I suspect you will also have a major battle on your hands if you try this, but
that's neither here nor there.

But until and unless this is done, the only appropriate thing to put in the
document at hand is ABNF aligned with what the standards currently say and the
planned revisions to these standards say. And this is what is in there now.
It is absolutely not the right place to change the set of characters in an
domain name used in email.

> I accept I could be wrong, but I do believe I have reasonable grounds for
> the view I have expressed, and if the truth is something different then I
> think the IETF needs to take steps to clearly document the true situation.

I'm afraid I have to agree.

				Ned