RE: RDF in OO-databases

Hi Mark,

Mark said:
Only just spotted this - but I agree with you Didier. I've actually
implemented a system using Microsoft's LDAP server to store XML nodes and
RDF nodes. I'm not yet on Win2000 so I don't know if the performance will
improve when I switch (and also use SQL 7), but at the moment it is not too
good. However, the advantages of the ease with which I can change the meta
data coupled with the resilience of LDAP have made me stick with it.

I think the next piece of the jigsaw is the interface. I have written (or
shall we say - am still writing!) a WebDAV layer over the LDAP. So far I can
browse my XML with Microsoft's WebFolder client, and upload NITF documents
which get broken into XML nodes, and then have loads of RDF added to them.

Didier reply:
I agree with you, going through LDAP to obtain a node is not fast. But it is
faster if the converter and the server is on the same machine.
Here is what I do and what are my goals:

The goal is to provide LDAP content as XML to the outside world. My first
try is to use RDF since the frame based pattern used by RDF seems to fit the
frame based pattern of directory services.

thus, a browser can do a request to an LDAP server and get back directory
content through LDAP protocol. In that case, an LDAP protocol handler is
necessary to convert the data received from the directory server into
something displayable. We have actually implemented an LDAP protocol handler
that works with IE 5. The protocol handler convert the LDAP data into RDF
elements and these latter and further transformed into HTML by an XSLT style
sheet. Thus, if the user do a view source, the XML formatted RDF content is
displayed and, in the document viewer, the RDF elements are transformed into
HTML and hence transformed into rendition objects.

On the other side, the goal is to have also the directory data available as
RDF documents under the HTTP protocol that, as you know, can pass through
firewalls. In that case, the following query:
http://www.netfolder.com/LDAP?O=W3C,C=US returns an XML document and more
particular an RDF document containing about his

<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl"
href="http://www.netfolder.com/LDAP/style.xsl" media="screen"?>
<rdf:RDF>
    <xml:base href="http://www.netfolder.com/LDAP"/>

    <rdf:description about="O=W3C,C=US">
        frame content here....
    </rdf:description>
</rdf:RDF>

We added the xml:base element so that each description is relative to the
base. this way it is much easier to integrate the description into a new
hierarchy like for instance:
C=Country
    |____ S=State
                |____ O=Organization

In the original taxonomy, W3C was classified as:
C=Country
    |____ S=State

To include the base in each description URI would impose to the RDF
processor to extract the domain to only keep the important information (i.e.
the element classification). Using the xml:base element we found that (a)
each resource is not polluted by location information, (b) the description
processing or inclusion into a new taxonomy is facilitated, (c) it is a lot
easier to read also and more semantically significant.

Our next step is that the server may recognize the user agent and do the
XML -> HTML transformation client side or server side based on a client
classification schema. For instance, a Netscape 4.x browser that do not have
a clue of what XML is, the RDF document is transformed server side. if a
client is an application, then the RDF document is delivered as is, idem if
the browser is, for instance, IE 5.x.

We found that this scheme works if an HTTP server is installed on the same
physical server as the LDAP server. Then, the original LDAP server gets
augmented by new capabilities by allowing access as usual through LDAP and
also now through XML.

By the way, if anybody is interested to test and play this new LDAP/RDF
server, let me know, as soon as the alpha tests are finished, I can ship to
you a copy to play with.

On the other hand, I am producing a document about LDAP to RDF mapping, I
will need review and constructive comments on the document. All
collaborators name will be (with their consent) be included in the document.
This could be, a good example of a practical application of RDF. And strange
enough, directory services are meta data containers :-))

Didier PH Martin
----------------------------------------------
Email: martind@netfolder.com
Conferences: Web New York (http://www.mfweb.com)
Book to come soon: XML Pro published by Wrox Press
Products: http://www.netfolder.com

Received on Tuesday, 18 January 2000 11:18:10 UTC