Re: DNS and Directory Services

["Martin J. Duerst" <mduerst@ifi.unizh.ch>]

Some observations re. DNS as a directory service, with
respect to several recent contributions from Simon Higgs,
Paul A. Vixie, Michael Dillon, and others:


The fact that [certain parts of] domain names bear a certain
resemblance to names in the "rest of the world" (and is not
just a random number) has several functions:

- Memorizability: It's easier to remember a domain name if it
	has some relation to the general name of the entity	
	concerned. It also helps in detecting spelling errors
	and similar phenomena.

- Understandability: If you see a domain name, you can understand,
	or at least guess, whom it belongs to, esp. if you already
	know the entity it belongs to.

- Constructability: If you know the name of the entity, you might
	be able to construct its domain name. This is what makes
	DNS a [poor] directory service. As has been pointed out,
	this is strengthened by the fact that people expect it
	to be so. It is also strengthened by the fact that the
	effort to try it is rather low, compared to other ways
	of finding/checking a domain name, and by the fact that
	people develop some kind of instinct to distinguish between
	those cases where it might work and where not.

- Identification: Older systems used numbers for user ids.
	People seem to prefer something close to their name,
	even if they have to make compromises if e.g. their
	name is John Smith. The same applies to companies and
	other entities. Numbers in the telephone system are 
	tolerated because of the historical hardware limitations.

It is important that the use of DNS as a kind of directory service
is only one of at least four reasons for why domain names are not
random numbers.



There are various reasons why there are no "real" directory
services widely deployed. Basically, they are squeezed from
below, from above, and from the side.

- From below, DNS is providing a poor but in many cases sufficient
	"directory service".

- From the side, the large number of concurrencing proposals and
	implementations did not (yet?) provide a single winner,
	an important requirement for an internet directory service.

- From above, the Web search services provide a kind of directory
	service. This service is again poor, but it complements
	well with what DNS is providing.

The experience with DNS and the Web also shows an additional
problem for directory services: How to collect the necessary
data. HINFO or TXT records could provide the necessary
data for a usable directory service atop DNS, but are very
rarely filled in. The Web search services collect their
data mostly with spiders/robots on the web, and not by
active contribution from the data source, as envisioned
for real directories.

Regards,	Martin.