Lists Home |
Date Index |
Howard Katz wrote:
> I don't understand this last point, Elliotte. How can a properly designed
> application ask whether a document contains the information it needs without
> knowing about the document's structure? If you add information, you're most
> likely changing the structure, and consequently the schema. How can an
> application cope with ad hoc changes like that w/out looking at the schema,
> ie without doing validation?
It's not hard if the added information is in the form of additional
child elements or additional attributes to existing elements. When you
ask for the specific child elements that you want, you simply will not
get the new ones.
Of course, if you go rearranging the basic containment structure, it's a
different ball game.
RDF is the ultimate along these lines - though not the xml/rdf syntax -
because there is really only one structure, so there is never anything
At least, that is the argument. I think that it is true in a somewhat
superficial sense. I think that a lot of real-world complex RDF will
make use of what I tend to call "idioms" - these are particular
groupings that will be found repeatedly. For example, if you translate
a set of relational tables to RDF you will get repeated and
characteristic structures (or subgraphs). These are the idioms.
Processors will probably need to understand the idioms in use to get the
most out of these rdf data sets.