John,
>
>Len,
>
>That was the major motivation for moving from file systems to database
>systems in the 1960s and to relational DB systems in the 1970s:
>
I am sure you will agree that when dealing with ontology - implicit in "scale" should be openness. If DBs were sufficiently open we would not need Semantic Web.
>LY> How do you maintain a body of knowledge useful and accessible to
> > many different application (since cost of developing a general
> > knowledge is not going to be justifiable for a single application)?
>
>When you bring in the WWW, the problem escalates. But it's important
>to recognize that large enterprises had world-wide intranets long
>before Arpanet became the Internet. DB systems, both relational
>and object-oriented, had been processing terabytes and petabytes
>when the Semantic Web was just getting started.
>
Undoubtedly, DBs are order of magnitude more efficient then SW. But they always struggled with distributed data processing.
>LY> Far from defending SemWebbers - I don't think they are ignorant
> > crowd.
>
>I agree that there are many very intelligent researchers working
>on the Semantic Web. But I believe that they could make bigger
>*and* better contributions if they were working independently
>rather than in a huge committee. Too many cooks spoil the broth.
>
>Note that Google was started by a very small group at Stanford
>at around the same time that Tim B-L published his book. But
>Google grew much bigger and much faster than the SemWeb -- and
>their technology is much more scalable.
Google is notoriously closed system when it comes to integration with other business applications. Google Base never delivered on expectations of open distributed database. They dealt brilliantly with distributed data processing challenge in a closed architecture (does anyone have any idea how they do it?).
>
>LY> To begin with - sW technology is primarily about Web (sorry for
> > offending anyone with small "s"). But it does address the main
> > challenge that is in my view -scalability. That does not disqualify
> > it from achieving "S" in some cases. In fact they are very good at
> > identifying these cases and dividing the problem space into specific
> > areas with different performance characteristics. So I would start
> > with looking at SW literature for technical incites.
>
>The SW has the strengths and weaknesses of a large committee. Instead
>of a clean, elegant, streamlined solution, you get a compromise that
>includes everything plus the kitchen sink. As an example, just look
>at the RDF notation for triples compared to the notation that Google
>uses: {A, B, C}. That is the JSON notation, which also supports
>n-tuples, objects, and typed or untyped terms and lists.
I don't know any good query language (not to mention manipulation) for Google or JSON data. And even though SW is far behind DB on that, at least you can build a simple application on the Web scale with it. Can you suggest a better way of doing it?
>
>I think that *ideology* is the main obstacle that has strangled
>innovation in the SW. If anybody whispers that JSON might be better
>than RDF, the SW thought police immediately exile them from the empire.
>
I agree, but what is an alternative? What does Google offer in a way of data integration and business application development? I think Microsoft offers more, as much as evrybody wish that someone replaces it with open solution.
And "cloud computing" is just... cloud at this point. So where do you go in not SW?
>JFS>> Protégé is useful for generating OWL, but OWL is not integrated
> >> with what business software developers, vendors, and the people
> >> who develop enterprise data bases and knowledge bases actually do.
>
>LY> That is true about Protege v4 which is based on OWL. However,
> > Protege is designed to be highly extensible and flexible. It
> > definitely can support enterprise applications and can be scaled up
> > significantly. Protege versions before v4 were based entirely on
> > frames (not OWL). I used Protege 3 to integrate it with data bases
> > for enterprise application development. You can download the OntoBase
> > plug-in I developed for that
http://www.ontospace.net/ontobase
>
>Protege is a good system, and the Stanford group should be given a lot
>of credit for developing such a widely used tool. But just compare
>two groups that both started at Stanford around the same time:
>
> 1. Google became a multi-billion dollar enterprise, which indexes
> all the publicly accessible web pages on the Internet. They have
> professionally developed tools that scale to all that data. When
> they combined asynchronous _javascript_ with XML to support Google
> Maps, developers around the world immediately adopted it, renamed
> it AJAX, and made it the foundation for thousands of web sites.
>
> 2. Protege is still a small group with software developed by student
> labor, and it depends on free plug-ins written by individuals
> who want to apply it to business applications.
>
>JFS> If you want to support business applications, you have to work
> >> with the people who develop business applications. And you have
> >> to make sure that your tools are integrated with their tools.
>
>LY> And Protege is still the best tool for that.
>
>As long as a university project remains the primary tool, you don't
>have a scalable methodology that can support a major enterprise.
I agree. But again - what is the better tool?
>
>John