What names or addresses things eventually map to should be fairly irrelevant
in terms of perceived complexity, as you'd alias Joe Smith@whereever to some
short version in your personal alias/nickname file anyway.
I think we're in violent agreement. My old arguments against explicit paths
are starting to hold true for what we are currently calling "addresses". As
I said in my last note, mapping of real-world objects to funny-world objects
(that is, person/place names to host/network names) is a directory services
problem and cannot be solved with DNS or anything like DNS. We've been
getting away with using DNS because our total number of objects has been very
small (70,000 .COM domains ain't nothin' in the real world.)
Or is that just
mapping information for a data base supported translation server?
That's how X.500 planned to do it. I'm pretty sure the IETF has stuff going
on in the directory services arena, perhaps someone on that WG could comment.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Here I borth agree and disagree. It is fair to say that DNS is not an
appropriate solution for the problems, however I disagree that something
like DNS is also inappropriate. What DNS does is provide an arbitrary
mapping between an individual and something related to that individual,
namely a machine name at the place they do most of their net access. If
we simply broaden things so that the arbitrary mapping can be chosen by
the individual and the mapping can be portable, then I believe the
problem is solved.
By inserting another directory layer similar to DNS but with different
top-level domains we can broaden the namespace as much as is needed.
Suitable modifications could be made to tools like sendmail to query the
directory layer either before or after querying DNS. The directory layer
would then issue a DNS name for further lookup.
Therefore, VIXIE.BIND.HACKER.ROLE would resolve to VIXIE.SF.CA.US today
and if you move to Australia you need only change the directory entry to
map to VIXIE.COM.AU or whatever. Since we have 36 symbols to use for each
level of naming, we should attempt to allow as many combinations as
possible especially at the top levels.
The only reason I can see for using a different mapping layer is that the
DNS system has other jobs to do and we should not overburden it with huge
databases that do not help it carry out its primary function.
Michael Dillon Voice: +1-604-546-8022
Memra Software Inc. Fax: +1-604-542-4130
http://www.memra.com E-mail: michael@memra.com