Re: NADF and other pilots [ was Re: UK Academic...]

[Christian Huitema <[e-mail]>]

>   The NADF and the Internet also differ in that for the most
>   part (again, excluding France, sorry don't mean to pick on
>   the French), the Internet manages knowledge (DSA entries)
>   using the Directory itself, whereas the NADF manages Directory
>   knowledge by means outside of the Directory.
>
In fact, the French pilots do manage knowledge using the Directory
itself. The simple difference here with the Quipu pilots is that they dont
use the same attributes, and they dont store them at the same place:
knowledge information is entirely contained within the DSA entries, not
within the "non-leaf" entries. This is for historical reasons -- Pizarro was
designed before Quipu was released. Also the scheme is easier to administer:
you have to replicate the DSA entries in all cases; by inserting the
knowledge in this entry, you avoid the need for replicating the "non leaf"
itself.

To come back to Erik's first question on "how many pilots", I would
underline the very big difference between the "listing" and "registration"
model.

The registration model, currently in use within the DNS and the quipu
pilots, mostly say that you need one "master" per entry, and that you cannot
create a subordinate name (e.g. an org within a city) without registering
explicitely the subordinate entry within the "master" DSA (e.g. the DSA run
by the town hall for the city).

The listing model, advocated by the NADF pilot, observe that you cannot
accept to have only one master for objects like "countries", "regions" or
"cities". Governments and town halls dont seem very much interested in
providing the service, and are in fact forbidden to do it by
"anti-monopolistic" laws in some countries. The same anti-monopolistic laws
would in fact forbid any company to assert a nation-wide monopoly on the
naming service; thus, several companies will "compete" to provide "naming
information" to various classes of "customers".

Mix it with privacy laws, trade mark laws and a grain of European
regulations, and you obtain a model where:

* organisations get a name by some magic -- i.e. using a registration system
external to the directory. NADF proposes using the "civil registries", i.e.
the formal registration of orgs for "civil" purposes. One could imagine that
other registration systems like the Internet DNS or the various name
allocation systems provided by standard organizations could be used as well.

* the organisation can run its internal directory using X.500.

* it will also request to "be listed" in various servers.

If you want to know something on the organisation, you will have to go to
one of these servers, which all have their own "view of the world". You will
probably be given back some informations + perhaps the address of the
organisation's DSA, and will be able to navigate towards the destination.

Indeed, this is different from the original "strict hierarchy" model of
X.500. At the top, we dont have a tree anymore, but a graph. Graphs and
tree dont mix well, which explain the NADF "isolation" from the quipu
pilots. And it is not at all clear that the X.500 navigation concepts remain
adequate!

Christian Huitema