Re: DNS under o=Internet

[[e-mail]]

[Note to all: Stef and I are in violent agreement. This message is just
 a few more comments.]

> What we cannot do is require that these two always be combined.

Right: the listing model does not preclude some subtrees being operated
on a registration model. The converse is not true.

> First, the idea should be to simply graft the entire DNS tree under
> what ever arc you chose to hang it under.  NO REREGISTRATIONS!  It si
> done wholesale by simply registering the DNS tree under some arc.

Just to make one thing clear: as above, no reregistrations, but we
cannot just have the DNS tree in its entirety appear under o=Internet
magically either.

There is a privacy issue here: people have to have the option of
choosing not to be listed. So we cannot just take zone files
and load the information in. This unfortunately somewhat curtails
the effectiveness of the DNS tree (anything short of the whole DNS
tree would), but unless the domain registrars want to create a
(retroactive) rule that registration implies permission to
have a Directory operator list the domain registered, there isn't
a whole lot we can do about it without seriously stepping on
people's toes.

Given the absence of the rule I mentioned, I believe that only the owner
of the domain should be allowed to load zone information for that domain
in.

So, for now, a DNS registration is a necessary, but not sufficient condition
for someone's subtree to show up in the portion of the DIT holding
the DNS subtree.

> But, I have a question about this issue: Where is o=internet
> registered?  I have not seen anything other than the idea of adding it
> under the root,as a peer of the ISO-3166 registered countries.

That was the idea. Since o=Internet is a listing point, the issue
of whether we can register the Internet at the root never comes up:
it's just an artifact that Directory operators use to group things
together.

Would it be better if we renamed o=Internet to cn=Internet?
A case can certainly be made that the Internet is not an
organization in any conventional sense of the word.  With the
consensus of the group, we can probably define some trivial object
class for that node.  This would take care of "what if some
organization trademarked the Internet name throughout the world",
since even in that worst case, the organization would have the right
to o=Internet, not cn=Internet.  Does anyone want to do that? I
hesitate to advocate this only because there is no effective
difference ("that which we call a rose ..." :-)) between the two,
and it would be a very artificial object class that got created.
Let me note however that if we want to do this, now's the best time,
before we start populating the tree heavily and have to be backwards
compatible.


Wengyik