Semantic Web (Was: [ANN]: /refind.xml)

[Daniel Veillard <veillard@redhat.com>]

On Thu, Aug 02, 2001 at 11:45:10AM -0400, team refind wrote:
> The semantic web will never happen until all members of society
> participate. People will *never* participate until they see benefits. The
> burden is on us all as developers to make things drastically simpler and
> empower every citizen with the power of information. Only when computers can

 Hum, semantic: yes, web: yes ! semantic web: very doubtful ...

 After 3 years of rpmfind experience (large catalog/database of mostly 
Linux free software [1]), well I don't believe much in a Semantic Web as
defined by most people. I do provide an RDF [2] catalog for all the database
content. I think I can try to draw a few lessons from this experiment up to
now:

   - RDF serialization is so scary that other people (like Ximian [3]) which
     needed similar package description used another format
   - having the data exposed on the web, actually led to some reuse, though
     is seems that for such databases people prefers to grab a snapshot,
     and work on it, not query the web each time
   - trying to make the data format "self explanatory" didn't help, 
     the various code base using the rpmfind RDF did not merge, even if
     I tried to provide a small module in the rpm2html/rpmfind code
     isolating access methods. I never noticed any generic RDF engine
     crawling the database (but I have seen classic XML indexing engines)
   - it didn't solved the problem of aggregation which is the most annoying
     one for catalogs (you build one, I build one, how does someone else
     query the full set)
   - XML didn't make it in the browser, so the XML database is imited to
     tools consumption, and cannot be browsed directly by humans. The
     advantage of using a portable format didn't really scale beyond
     specialized tools. On the other hand the web database front-end
     has been a great success (many Mhits/month).
   - Initially the XML format was the database and HTML pages were generated
     from them. At some point I really had to use a database engine to allow
     powerful/fast searches, now the RDF is generated from the database
     like the HTML pages.

 Basically the result is mitigated, yes exposing the XML database on the
web was useful for tools. But it didn't reach humans. Hence it did stay
limited to the audience using the specific tools, and the hopes that the
genericity of a format like RDF would allow unspecialized engines to profit
from it didn't realized until now.

 If I had to rebuild the service from scratch I would probably still use it
anyways, maybe the next report 3 years from now will allow to claim success,
but it is really doubtful. I'm also  still waiting for a free software
database targetted at XML, I'm more optimistic on this one :-)

Daniel

[1] http://rpmfind.net/linux/RPM/
[2] http://rpmfind.net/linux/RDF/
[3] http://ximian.com/

-- 
Daniel Veillard      | Red Hat Network http://redhat.com/products/network/
veillard@redhat.com  | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/