The DNS is not a directory

[Dave Crocker <dcrocker at brandenburg.com>]

At 10:18 AM 4/23/98 -0400, Peter Deutsch wrote:
>DNS is currently used as a directory system. To quote the
>line from E.T. "this is reality, Greg". You may not like

Peter,

The DNS is used less as a directory system than current mythology claims.
In fact, I claim it isn't used as a directory system at all.

The fact that, for example, you can use the name of my company and guess my
web address does encourage people to think of the DNS as a directory
service -- i.e., as a mechanism for doing searching -- but it is
technically wrong and it fails frequently.  In addition, the limited extent
to which people believe their current use of the DNS is the same as a
directory service is based on a phenomenon which doesn't scale.

Here's why:

1.  The "searching" is done by the user (or the user's client software),
not the DNS.  The user guesses a specific domain name.  If they guess
correctly, they win.  If they guess incorrectly, they get to guess again.
In contrast with a true directory service, they are never given a list of
multiple hits from which to choose.

2.  Most of the successful pseudo-directory use of the DNS today is
predicated on having the domain name end in .com.  (If you know enough
about the organization you can sometimes guess .edu or .org correctly; it
is very rare to guess .net correctly and even more rare to choose between
.net and .com, when there is the same second-level domain under both.)
Besides the high rate of failure for the .com "heuristic" just within the
set of gTLDs, it ignores the 200+ ccTLDs, much to the irritation of the
rest of the world.  (There is an IMAP email client, called Simeon, made by
Esys.  Quick.  Guess their domain name.)

3.  As the number of domain names grows, one or both of two phenomena
happen:  For a new registrant the name that is most appropriate to choose
is unfortunately already taken, so now you must choose one that is less
appropriate.  That ALWAYS means harder to guess.  And/or adding top-level
domains makes guessing the right TLD impossible.

d/

ps.  As I've commented to you privately, I think that the whois++ concept
is quite wonderful, since it allows laying a relational data model on top
of a hierarchical (and other) storage mechanism.  That is exactly what the
DNS is, and so it would be a very interesting exercise, indeed, to develop
a whois++ layer on top of the DNS.  That new layer very much WOULD BE a
directory service.  Even better, this added functionality can be created
incrementally, without disturbing the installed base of data or service.
That's would be an stunning accomplishment and exactly what the Internet
loves, since we like to protect the installed base.
__________________________________________________________________________
Dave Crocker                 Brandenburg Consulting        +1 408 246 8253
dcrocker at brandenburg.com       675 Spruce Drive        (f) +1 408 273 6464
www.brandenburg.com         Sunnyvale, CA 94086  USA