Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!umd5!brl-adm!adm!Lennart_Lovstrand.EuroPARC@Xerox.COM
From: Lennart_Lovstrand.EuroPARC@Xerox.COM
Newsgroups: comp.mail.headers
Subject: Re: ! and @
Message-ID: <12868@brl-adm.ARPA>
Date: 8 Apr 88 11:05:33 GMT
Sender: news@brl-adm.ARPA
Lines: 30

Sorry, that won't work either.  Two examples:

1) Localhost is connected to two machines 'unics' and 'relayhost', but does not
have access to any name server.  It does, however, know that 'relayhost' can
take care of most unknown addresses.  In comes now the following address:
	unics!gobblygook@acme.com
Should this be sent to the local 'unics' just because there happened to be one
in the neighborhood?  I say no, it is just as plausible that acme.com also has a
local machine called 'unics' and that the message should be sent to 'relayhost'.

2) Localhost is both on the Internet and the UUCP network and has access to
bind.  In comes the following addresses:
	rambo.gov!gobblygook@state.edu
Both rambo.gov and state.edu are on the Internet and are therefore known to
localhost; to which destination should it send the message?

It's not even safe to use different parsing strategies depending on from what
network the message came, as UUCP more and more is adopting domain style
addressing.

The answer is that there is no patent proof way of disambiguating the mixed !/@
syntax.  Heuristics may help, yes, but as long as these hybrid addresses are
being used there will always be confusion and misunderstanding by the people and
machines that have to deal with them.  Only by persuading one of the communities
to adopt the other's standard can we get rid of this pain -- or by scrapping
them both and universally start using something like X.400.  In the mean time, I
think it's better to keep a firm parsing policy which favors exactly one of the
above -- and if you're connected to the Internet, you know which one to choose.
:-)

--Lennart