what is a dimension ?

In a sense, this is a trivial question. It is just a degree of freedom, so that there are as many dimensions as there are independent variables, so in a dimensional model there will be a central FACT, and n orbiting dimensions – one for each independent feature of that fact. – but this definition just shifts the notion of dimension to that of independence, and while the notion of dimension is arguably an ontological question, the notion of independence is a logical and perhaps even an epistemological question. where does this leave us.

any particular solution is just a solution. It carves up the world in such and such a way, and it does it well or not so well according to what you plan to do with the result. This is not an anti-realist rave, but a realisation that any ontology is JUST a way of seeing, which ironically places ontology BELOW phenomenology not above it.

Ontology is NOT more basic, but it is just less immediate, and while descriptions of experience are not the same as descriptions of worlds, the distinction is perhaps not so decisive as the purpose of the subdivision.

All this amounts to my substantive claim that this exercise is precisely the identification of independent dimensions, and that a data model just expresses such a perspective. The real task, the really hard one, is working out what IS dependent on what, and what is independent – and how to encode that dependency in a representation of the subject domain that carves it up in a repeatable and objective way.

In the end, this is a modal distinction that a data model demonstrates through the distribution of primary and secondary keys and this is not a component OF the model but a WAY the model is. It is a MODE not a COMPONENT.

So what IS a dimension. IT is not a component, not a property, not a mode. – this has become quite an interesting problem….