Scaling. Your proposal makes every message go to thru one single machine,
bringing it to it's knees. Now or tomorrow. Or yester- day.
Not hardly. Many implementations may force the messages to go through one
single machine, but that is not inherent in the address. AT&T Mail supports
a fax addressing scheme similar to that shown above, and it definitely does
NOT use a single machine to do its fax deliveries.
Of course you can implement anything you want in any way you like. You can
route on the Fruit-of-the-day: header if it suits your fancy. But the existing
routing structure provided by the DNS and MX records is already in place.
Replacing or enhancing it would be a monumental undertaking if it could be done
at all (which I strongly doubt). So, given that:
(1) We have a system in place that represents billions of dollars of
(2) It isn't going to go away and isn't likely to be updated or updateable.
(3) It can be used, with no modifications, to give us the service we want.
(4) It imposes some restrictions on address format, but the restrictions are
entirely reasonable, and no other format will work.
There is only one rational choice, and it is the one Marshall and Carl made. (I
expect nothing less from engineers of their caliber.)
I also don't entirely agree with the evaluations of the relative intuitiveness
of the two formats. Specifically, neither of them are especially intuitive to
an unsophisticated user, and I don't think comparing the degree of
nonintuitiveness of two nonintuitive formats is a useful exercise. (As it
happens I have substantial experience with the address format you propose,
since my software uses something quite close to it.)
In either case a decent front end is urgently needed, and not just for
addresses -- cover page information usually has to be provided as well. Once
this is done the underlying address format ceases to be an issue aside from its
ability to interwork with Internet mail.