Re: [DotGNU]RDF based discovery system

[James Michael DuPont]

--- Peter Minten <address@hidden> wrote:
> James Michael DuPont wrote:
> > 
> > This ties back into my original proposal for a meta-object
> workgroup
> > and meta-data repository.
> > http://wiki.dotgnu.org/TaskMetaDataRepository
> > 
> > That would provide many datasources for a generic discovery system.
> 
> GNU.RDF is open enough to accept data from those sources.

OK, That is good news. We need to be able to provide RDF interfaces
into existing software. That would be best by being able to make a form
of gateway into existing data structures. A simple acccessor layer
would be implemented on top of a data structure that would allow the
data stored inside it to be converted into rdf. That is like
implementing a DotGNU.RDf.Parser interface.

The next step would be the implementation of a DotGNU.Rdf.Storage
interface on top of a data structure . That would allow for the
accessing of the data without having to convert all of it into RDF. For
a large data structure such as a parsed IL file this would be very
efficient. The tools that run RDF Queries should be able to be ported
to any tool that supports the DG.RDF.Storage interface.

>  
> > We need to really ask ourselves what is going to be published and
> what
> > not!
> 
> There will be a standard, but it can be extended.

Lets say, I am a business. I want to see the individual functions of
the linux kernel as webservices, or as source code. 

Each function would be documented, and the datatypes of the parameters
would also be cross referenced. 

You should be able run the following queries on the discovery system :

1.  What are the things that I need to use this function :
1.1 What are the parameters, what are the data types.
1.2 What are the fields of the data types, what functions create them,
what are the documentations for them.
2. What are the return values of the function, how can I deal with
them, what functions use them?
3. What are examples of the function, how can I use them? 

These are some of the basic discovery functions that will help people
decide what function they what to buy/invoke.

On the the other side, in terms of a discovery system, we can think of
business functions that also have data types. For example, we have a
AutoMechanic, he can repair cars of Type Volkswagen. 

We would like to be able to select all the people who provide repair
services in an area who are able to repair volkswagens.

>  
> > Many free software developers lack the creativity to think
> themselves
> > out of the box.
> 
> I fear that's true. A good sign is the languages used. C is the top
> language,
> not LISP. C is pretty strict and you can't really do new things with
> it because
> it's too low level. LISP is a hugely flexible language, but few use
> it (or have
> the guts to try to learn it).

Well, I would like to see some type of pnet interface into emacs and
lisp. The swig should provide some of that. 

>  
> > The idea of discovery needs to go down to the method and variable
> level
> > where you can browse an program and extract just the needed bits
> and
> > bobs. In the end, this is tied closely into the job of the
> compiler,
> > but is implemented on a global scale.
> 
> Introspect, extract into RDF, publish on an RDF server and make the
> standard
> project metadata link to it.

Ok, in that case, I would just be a vendor hawking my wares. Sure.

> 
> > I challenge you to have the courage to embrace such an idea.
> > 
> > I don't think it is possible because of the lack of knowledge
> exchange
> > and openness.
> 
> Hmm, I know of a tool that makes it possible :-).


:)

Thanks for your support peter.

mike

=====
James Michael DuPont
http://introspector.sourceforge.net/

__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com