RE: A philosophical question about large directory structures

Harald Tveit Alvestrand <[email protected]> Wed, 23 Jun 1999 15:18:50 +0200
Newsgroups gmane.ietf.lsd
Message-ID <[email protected]>
At 14:58 23.06.99 +0200, Peter Jurg wrote:

>Referrals are URL's and therefore always unique.

say that in the presence of someone who loves URNs (but wear an
asbestos suit :-)

>  So if you register at an referral server no name spaces are imposed.

au contraire - the name space of URLs (method, hostname if method is LDAP:) 
is imposed on the top of the name spaces used by DNs. The resulting 2-level 
namespace is more likely to be disjunct (no 2 objects have the same name at 
any instant in time), but more likely to be multiply populated (multiple 
names resolve to the same object).

>  Now what is the difference between registering according to the imposed 
> name spaces and registering to an ldap referral server? It is still a 
> matter of registering (again, without having a clear idea about what 
> registering should be). The difference would be that the distribution 
> according to a DIT is no longer possible.

And that DN-valued references between namespaces are impossible; they have 
to be URL-valued.

>  And probably one would need many more (i.e. more elaborated) ldap 
> referrals without imposed name spaces than with them. I don't think that 
> is a big deal. Many other possibilities are still open.

that's for sure!

>The focus IMHO should be on the registration (or discovery) of servers and 
>heir content. Not on the name spaces.

I think we need to talk about services, not servers.....and we have to name 
them somehow, if we want to get rid of "The Service" (which is unique, and 
therefore doesn't need a name).
The (DNS) name of their (main) server isn't a solution for naming services 
I feel good about.

>+ >The real problem is standardizing object classes and attributes so that
>+ >people can actually read information from each other's directories.
>+
>+ interesting - this came up > 1 time now....
>
>I noticed a rather extensive discussion on the ldap-ext list regarding the 
>top objectclass ;-)

at least there you can tell right from wrong :-)

                Harald


>Peter Jurg
>M&I/STELVIO

--
Harald Tveit Alvestrand, Maxware, Norway
[email protected]