On 10/4/05, Anne Thomas Manes <atmanes@g...> wrote:
> It was persistent while you were working on it. Then you disposed of it.
> If you don't like the word "persistent", then come up with a different term.
> Perhaps "active"? Whatever...

I still can't see any reason to try and label XML by how long it hangs
around?  (As such, I'm don't see value in trying to come up with such
a term.)

>
> Your XML won't care, but both you and your application probably should
> because the data model impacts performance and efficiency. Our core question
> here is focused on proper modeling of the XML structures -- whether they
> should be normalized or not.

I for one don't see much of a useful general connection between
temporal duration and XML models.  (Please no forays into the issue of
"6th normal form"; normalization for modelling temporal issues really
isn't what's being questioned here.)

I don't expect to see a lot of success trying to formulate broad
general rules for XML modelling.  Even within the relatively well
defined realm of RDB modelling one can normalize any reasonably
complex domain many different ways each of which has different
implications for data management and application design.

>
> If the XML is transient, then it probably ought to be normalized, because it
> offers the best optimization for transformation.

Don't agree. What if my XML is transient precisely because it is the
result of some previous transform that turned it into some form of
temporary XML (eg. XHTML, SVG, Docbook) perhaps starting from some
normalized form?  If anything I suspect the opposite to be true: I
like normalized relational databases; and they do tend to be
associated with some degree of persistence....

> If the XML is persistent, then application requirements ought to dictate the
> need for normalization.

Leave off the first 6 words and we're getting somewhere...

> If the XML will be operated on by only one
> application, then the structure probably ought to be optimized for that
> specific application.

Most definitely.

> If the XML will be operated on by more than one
> application, then the structure probably ought to be normalized.

Don't think there's a general rule here either.  What if all the app's
are Web browsers?  Should I denormalize my XHTML with embedded SVG
into something else?  I don't think so...

--
Peter Hunsberger