I would comment that the DNS is the Internet's namespace. So
tim(_at_)frobble(_dot_)org
is syntactic sugar for tim.frobble.org. It is good syntactic sugar because
it suggests that the named entity is a person, and it is a lot nicer if
there is a dot in the name to write tim(_dot_)smith(_at_)frobble(_dot_)org
than
"tim.smith".frobble.org .
I'm not sure that '@' is syntactic sugar for '.' since that causes
problems with users of the same name as subdomains. For example, I
can easily see a use named lcs in the mit.edu domain, which would
conflict with the lcs.mit.edu domain.
So, if '@' == '.', then lcs(_at_)mit(_dot_)edu -> lcs.mit.edu; how do you
differentiate the user from the domain?
Instead, you might want to think of '@' more in the sense of a Hesiod
meaning, you have a key to a key-value pair in some domain, and you
have special "maps" to map the keys to values. The hesiod method is
to use the format "<key>.<map>.ns.<domain>", so theoretically user lcs
would have something like:
lcs.cert.ns.mit.edu
And this would be very different than the domain "lcs.mit.edu", which
could theoretically contain "lcs.cert.ns.lcs.mit.edu" as well.
However the main thrust of your argument is probably sound, even if
the details might have been problematical.
-derek