Re: About the creation of new 3-letter domain(s)

[Robert Elz <kre@munnari.OZ.AU>]

    Date:        Wed, 18 May 94 15:00:17 -0700
    From:        Dave Crocker <dcrocker@Mordor.Stanford.EDU>
    Message-ID:  <199405182200.PAA00159@Mordor.Stanford.EDU>

    You can't do better than self-assignment.

No, but since the DNS is a tree, something higher up the
tree has to link you in.
    
    To the extent that names are self-assigning, there is a
    reasonable chance that the parent DNS nodes can automate the
    registration.

Apart from this clearly being future vision (the tools to
verify that what is being requested is anything like rational
don't exist yet), I still don't see the connection between
the value of the name chosen (ie: the contents of the string)
and the method by which its registered.   If there existed
some service at node a.b.c that I could contact and register
myself in its DNS server, I don't see it matters just where in
the domain tree it has positioned itself.   In all cases it
has to check that the name I am requesting isn't already
allocated (it may be easy for you to resolve two registrations
of "Crocker" from your house, but you can't expect software to
figure it out on any basis than first come).  Having done that
all it needs do is check the technicalities (are there servers
running where I say there are, and do they answer queries for
the name I want delegated, etc).

What have I missed?

    Randomness maximizes registration overhead.

Why?   What is different?   In a sense, randomness could make
things easier, as rather than having to work out that all the
other houses on my street have registered as mlk-way where I
wanted to have it shown as "Martin-Luther-King-Junior-Way"
(or even Grove St) randomness could be done by providing a
server somewhere that I mail to, which sends me back a pair
of random numbers (or I could run that as a local process).
Given the numbers, all I need is a quick DNS tree walk to find
the location of the server with which I should register, just
the same as is needed for a geographic based name.

    The more random, the less mnemonic.

That's certainly true.   Its also a semi-feature, as it
encourages more meticulous record keeping, almost no-one
is going to try to remember many random number pairs.  On
the other hand, addresses that look as if they can be remembered
can cause problems (is it "hair-st" or "hare-st"?  You hear it
spoken to you and one of those registers in your mind, the
fact that its the wrong one never occurs to either party).

    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.

Really?   How does the post office deliver mail without a
zip code in the US?   I know they do, as I almost never know
anyone's zip code, so just send mail without it.   I assume
I'm not alone.   If you tell a taxi driver to take you to some
address, do you include the zip code there as well?

    (Some of us remember changing from mnemonic telephone exchanges to
    numeric ones.  Harder to remember and not nearly as personable.)

We went through that too, I just barely recall having a phone
number with an alpha prefix, but ours had no mnemonic meaning
the way they seemed to in the US, ie: no-one ever associated
names with the letters the way it was done in the US, they
were simply letters, and hardly missed at all.   I recall being
(or my parents being more correctly) XY-1267 at one early stage
of my life, but it was just that, not "xylophone" or anything
strange like that.  "90-1267" was no harder to remember.

kre