RE: [xml-dev] Re[2]: SQL instead of XQuery [offtopic]

["Michael Kay" <mike@saxonica.com>]

> MK> cases where it works well:
> MK> (a) where the documents map well to the business objects that are 
> MK> the primary information content of the database.
> 
> What is the business objects ?

Sometimes it's obvious ("hotels"). Sometimes it's far from obvious ("rail
journeys"). That's why analysis is such fun. I've found that XML databases
work best in the case where there is a business object such as a hotel or a
plant species, where it doesn't take much analysis to understand what a
hotel (or a plant species) is, but where the data that you want to record
about each object is highly complex and variable.

> If documents are maping well (all items are predictable, 
> repeated), whether it's better to save under usual relational 
> engine instead to save xml ?

Frankly, I'm not very interested in data that's predictable. It doesn't
happen in real life, only in student textbooks. We use relational databases
when we can pretend that the data is predictable. We force the data to fit
into rows and columns by ignoring the parts of the information that can't
readily be modelled that way, and we pretend that it isn't changing by
ignoring the changes, or the requirements for change.

> MK> When these conditions don't hold, for example with an HR 
> database ...
> MK> document-centred modelling certainly has its limitations. ...
> 
> What limitations ?

If you've got a complex graph of objects, such as those that make up a
railway timetable, then there is no obvious or unique way of representing
the information as a set of documents. Two people doing the modelling will
come up with completely different solutions. In this situation, I think that
it's best to avoid doing the modelling in document terms, and the arguments
for storing the data in XML are not strong.

Michael Kay
http://www.saxonica.com/