[12471] in Commercialization & Privatization of the Internet
Re: About the creation of new 3-letter domain(s)
daemon@ATHENA.MIT.EDU (Dave Crocker)
Wed May 18 19:10:42 1994
To: Robert Elz <kre@munnari.oz.au>
Cc: bigz@rice.edu, namedroppers@rs.internic.net, com-priv@psi.com
Cc: dcrocker@mordor.stanford.edu
In-Reply-To: Your message of Thu, 19 May 94 07:16:43 +1000.
<18984.769295803@munnari.OZ.AU>
Date: Wed, 18 May 94 15:00:17 -0700
From: Dave Crocker <dcrocker@mordor.stanford.edu>
---- Included message:
However I fail to see what any of this has to do with the
administration of the DNS, and all the work that's involved
Well, since you asked:
The DNS consists of:
1. Name assignment
2. Server administration
3. Server access
4. General human use of DNS strings
In every case, the DNS has been designed to support distribution. This
includes name assignment, as well as server operation. Delegation of
name assignment was as key an enhancement to the net as was the
distributed storage/access service. The idea behind it was to reduce
the difficulties of getting a 'unique' name. I'm only trying to take
that concern to an extreme which is optimally workable for a truly
global service. You can't do better than self-assignment.
(as an aside, I'll comment that like everyone else, I'd love a unique
identifier for life. However, we don't have much experience with such
a beast, so that requiring it for any near-term service operation -- e.g.,
making it a core and required part of Internet expansion -- strikes me
as an example of tying together multiple dependencies and therefore
increasing the risk of, ummm, failure.)
there. Having picked a name, be it foo.com, or foo.123.456.pvm,
foo.16.main-st.hometown.dc.us (or something), it still has
to be registered, which involves finding the registrar (often
the hardest part for many people when the answer isn't
internic.net, or sri-nic.arpa...), and then getting the
registration made.
Fair point. To the extent that names are self-assigning, there is a
reasonable chance that the parent DNS nodes can automate the
registration. (You don't re-register when you move into a new home,
though you do have the option of registering it as a forwarding address
for the OLD one; but that's not required.) To the extent that names are
random, they can't. Randomness maximizes registration overhead.
Incidentally, there's nothing technically wrong with geographic
addresses, but they do cause people to try to read more into domain
names than should be read into them. The random numbers scheme has
the nice feature that its clearly meaningless.
The more random, the less mnemonic.
All the domain name should do is allow the correct DNS server
to be located to provide information about the name.
That's the computer requirement. I'm also trying to attend to some
human use issues.
ps: if we had geographic names with lots of detail, as you
suggested, I fail to see the use for the postal (zip) code.
That's a post office artifact to make their mail routing
easier, and adds nothing meaningful to the address at all.
I don't like postal codes either. But they seem to be the only way
to get the necessary granularity, given that street addresses are
allowed to be ambiguous within a city. And we are already used
to dealing with them. Sigh.
(Some of us remember changing from mnemonic telephone exchanges to
numeric ones. Harder to remember and not nearly as personable.)
Dave
P.S. Unfortunately, I am highly sympathetic to Christian's concerns about
monopolistic registration services. I've heard various stories about
problems in the current DNS world, so I assume it will only get worse.
Is there anything we can do about this?
D/
