Complete rewrite of the data model

[Tpt]

Note: this specification is intentionally not complete

[Ezibenroc]

I find it too formal. I prefer this kind of things:

union

∪: list × list × ... × list → list Denotes the concatenation of the lists. Example: ["Douglas Adams","Peru"] ∪ ["Peru"] ∪ ["22/11/2014"] = ["Douglas Adams","Peru,"Peru","22/11/2014"]

[Ezibenroc]

Notice the duplication of the element "Peru" in my example, the purpose is to remove ambiguousness: yes duplicate elements are allowed. We use set operators on lists, it may be confusing for the reader.

[Tpt]

@Ezibenroc Thank you for all these typo fix.

[yhamoudi]

please, could you add the possibility to have connectors of arity 1 such as First, Biggest, ...

[Tpt]

@yhamoudi Yes, of course. But I would prefer to add them in an other pull request, after we agree on the basic data model.

[Ezibenroc]

"something in the universe" was volunteer in order to say it may be everything (and it is the term use in RDF specification). "some basic object" is, I think, less clear and the word "basic" seems restrictive.

Ok

I linked the "true" as example in order to show that the two booleans are also resources

This functional notation is very odd. That mean that a predicate has simultaneously 2 types: list*list-> bool and list. This is pretty hard to unify. I prefer a curryfied notation to transform a predicate into his caracteristic function (what is needed here).

If true and false can be a resource and a boolean indiferently, then I don't see why a predicate could not be a function and a list.

Moreover, I really see predicates more as function than as lists (in fact, exactly as the notion of predicate in math).

[marc-chevalier]

Because booleans are a subset of resources and because function on lists are a higher order than lists.

[Ezibenroc]

I still think that a predicate is more a function list*list->bool than a list.

[Tpt]

A first point that, I think, is fundamental: a resource may represent anything, including a predicate, the conjunction or even a triple.

After discussions with Marc and reflexion I don't like too the idea to see in the data model predicates as functions of list_list -> bool (or resource_resource -> bool) and, so, the predicate(subject, object) notation.

Some reasons:

• Predicates are not only functions but functions with a semantic. We can imagine that there are two predicates p1 and p2 such that forall a b, p1(a,b) <-> p2(a, b) but p1 and p2 has two different meanings (like (it's a silly example) "son of", "got some genetic material from") and, so, are different. My opinion is that see them as functions is too much restrictive.
• I would like to clearly separate resources (e.g. things we talk about) and data model structures (like triples, conjunctions and so on). So, I don't think that see predicates as both resources and data model structure is a good idea. I believe that see only predicates as resources is far more simple and clear (but it's only a personal opinion).