Sunday, April 20, 2008

The Two Laws Of Dimensional Ontology

The Two Laws Of Dimensional Ontology

  • Law I
    • "One and the same phenomenon projected out of its own dimension into dimensions lower than its own is depicted in such a way that the individual pictures contradict one another."
  • Law II
    • "Different phenomena projected out of their own dimension into one dimension lower than their own are depicted in such a manner that the pictures are ambiguous."
I think this is two laws somehow explain what happens when we try to use a model to describe stuff in the real world, we end up representing things inaccurately and incompletely... the funny things that sometimes we seem to somehow change our angle of perspective, and then we realize that the our model is wrong... and try to fix it, and after that we believe that now it is right, but we will never have it right because the map is not the territory. Perhaps a similar problem arises when we try to use ObjectRelationalMapping, we simply lose some of the information because of the perspective we choose to use... Object Orientation forces a particular structure it may not exists... and Relational... trying to see things from a neutral point of view end up not being helpful enough: it seems that humans tend to need to see things from a particular hierarchical perspective at a time, and find it hard to grasp things from all points of view at the same time.

I started to think about this, because as a software engineer I am sometimes asked to create documentation about the software I am building (you know, the typical use cases stuff), and when you have to write a document of the kind a lot of times, the company where you work typically already have some document templates... now, typical document templates have a hierarchical (object oriented?) structure:
  • UseCases
    • UseCase1
      • Scenario 1
      • Scenario 2
      • Actors that participate in the use case
      • Classes (sterotyped as boundary in RobustnessAnalysis) that participate in the use case
      • Classes (sterotyped as control in RobustnessAnalysis) that participate in the use case
      • Classes (sterotyped as entities in RobustnessAnalysis) that participate in the use case
      • ActivityDiagram?
But then, they realize that some times, (for example to decide what uses cases will be affected by the modification of an entity class) they also need a document like this:
  • Classes
      • Classes (stereotyped as boundary in RobustnessAnalysis) that participate in the use case
        • Class 1
          • Use case 1
          • Use case 2
          • Actors interacting with the class
        • Class 2
          • etc,etc.
Both documents have kind of the same internal model, but represent (view it?) form a different perspective, now, the strange thing, is that I do not know of a commecially successful desktop publishing and/or word processing application that allows you to specify such a model so that you can describe your system once and only once and produce documents with several perspectives (I know products like Adobe FrameMaker are supposed to help you with this kind of semantical document structure, but, since they are (AFAIK) internally based on hierarchies, and not relational theory, they do not really allow the user to use all the power of the relational manipulation of data) I think that the problem might be that humans usually organize their ideas following some kind of hierarchical structure... or at least thats the way we are thought to do it at school, perhaps because of the limitations of printed material (a book made of paper can not be queried and reshaped relationally). But, now that we have all this electronic power to help us, we still try to organize stuff in hierarchies (or in messy graphs) and not relationally (internet is a messy graph, no a relational database... am I wrong? XHTML/XML are hierarchical, nobody would accuse them of being relational). We simply find it easier to organize stuff in that way... is it really because that is the natural way in which we think? or is it because we still do not have good enough relational tools to deal with knowledge? is relationally actually about creating a model that is valid from any perspective? or it is more about easily changing the perspective? (But if we build those model from particular perspective, how can we expect relational to produce a model that is valid for more than one perspective?) And finally, why we do not have any Relational Word Processor/ Relational Document Standard

1 comment:

Maxetormer said...

Hola, gracias por lo comentarios y links :)