Re: [xml-dev] Make implicit structures explicit

[Frank Manola <fmanola@acm.org>]

I'm talking about both logical and physical design.  Of course you're right to point out the differences between them, but the decision as to whether the design is going to be logical or physical is itself a distinction in how the design will be used.  Even when considering only conceptual design, there are potential design issues based on "usage" considerations *at that level*.  Your point about scoping is one of these, but there are others.  For example, there's been lots of work on ontologies for modeling spatial and temporal phenomena, and there are a number of different ways of "representing" (at a logical level) those things.  Choices between them are often made on the basis of personal preferences (are you a three-dimensionalist or a four-dimensionalist?), but sometimes it's just easier to model the things you need to model using one approach vs. another.  These choices don't always involve changes in the language used to express the model (in the case of ontologies it may all be first-order logic), but it does affect the primitive concepts used to develop the models within that language.   Ease of expressing logical-level mappings between a conceptual schema and views (also logical level) is another example that comes to mind (although this may fit under your "scoping" exception).  It's not just XML that's fuzzy as between logical and physical.  A given language may be better suited to logical vs. physical modeling, but that doesn't prevent someone from using it for either purpose, or a design in a "logical" language from being influenced by a lot of what are really "physical" considerations (e.g., lots of relational database designs).    

On Apr 9, 2014, at 3:51 AM, Michael Kay <mike@saxonica.com> wrote:

> 
> On 8 Apr 2014, at 22:44, Frank Manola <fmanola@acm.org> wrote:
> 
>> I think designing data structures without any consideration of what they are going to be used for isn't really "design" at all.
>> 
> 
> 
> Are you talking here about logical design or physical design?
> 
> I've always thought that with logical design (conceptual data modelling) it's a very good idea to ignore the expected usage of the data as much as possible - with perhaps one exception, the scoping that decides whether particular information should be included in the model or not.
> 
> Usage considerations should only come in at the physical design stage.
> 
> Of course, one weakness with XML, especially as a database technology, is that there's not a clean enough separation between the two.
> 
> Michael Kay
> Saxonica