From: Russ Allbery (rra@stanford.edu)
Date: Tue Jan 16 2001 - 07:16:52 CST
Charles Lindsey <chl@clw.cs.man.ac.uk> writes:
> So our document will probably be the first to mention RFC 2606 and
> ".invalid" explicitly. The question then is, can we, or should we, say
> something like "User agents MUST/SHOULD NOT attempt to send email to the
> TLD '.invalid'". One could argue that genuine interoperability did arise
> there, and I think that is what Brad is asking for. Or do we just say,
> in the NOTE, that any attempt to send such email will likely lead to a
> refusal by the MTA.
I don't understand how anyone could possibly consider this to be an
interoperability problem. Perhaps people are using a different definition
of that term than I am.
If you try to send mail to a .invalid address, that action will fail. RFC
2606 guarantees that. The idea here appears to be to argue in favor of
failing early rather than failing late. That seems to me to clearly be a
user interface and QoI issue.
An interoperability problem, I thought, meant that if one did that one's
implementation would not be able to successfully talk to another
conforming implementation of the standard, or that something significantly
bad might happen. The *protocol* act of sending mail to a .invalid
address is outside of the scope of our standard, since we aren't the SMTP
folks, and apart from that an article with a .invalid From address acts
just like any other news article, so I don't see how this could be the
first issue. I don't consider generating a bounce message to be something
significantly bad when the alternative is to just generate another type of
error message, so that seems to rule out the second issue.
Could someone explain to me what I'm missing?
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>