[om-list] Re: Data models

[Luke Call]

Mark Butler wrote:
> I agree completely that we should avoid fixed data models in the 
> traditional sense.  However we do need a common ground level data model 
> (or meta-model)  or we have little hope for exchanging models between 
> systems.  We have discussed meta models in the past, but I for one am 
> much more prepared to have a coherent discussion these days, based on 
> what I have learned over the past few years.

If the DB info I sent isn't readable enough to see how close our models
are now, let me know. I haven't modeled certainty at all yet, but have
noted your comments for when I do. Or maybe I will put them in if our
models are otherwise close enough to merge from.

I also wonder if someday semantic web protocols will be useful as a more
generic, high-level comm. tool between such models.

> My current family history data model is very generic - it has Concepts, 
> Attributes, Relations, and Evaluations, where the latter three are 
> sub-classes of the first.  I still create sub classes to bind code to 
> various types of objects, but the data itself is stored at a very high 
> level rather than in hard coded sub class specific fields.  This turned 
> out to be very important for several critical algorithms I am developing 
> - the one for intelligent model merging being the foremost.

very cool.

> At some point I would like to write an interpreted language that works 
> against the object model at a much higher level than can be achieved in 
> either Java or C++.  Declarative and procedural aspects could easily be 
> implemented as part of the same language.

I wonder if Scheme could do that for you--I'm reading the part about
macros, but essentially I believe you define your own language
constructs with them. The language itself is so powerful as a way to
manipulate symbols, which in turn can represent about anything. Its
former lack of libraries is overcome by an implementation in Java, I
believe.

Luke