Lists Home |
Date Index |
> >2. propogation of text nodes up from elements in namespace y to
> >elements in namespace x, example:
> ><y:y>text node</y:y>
> >in this scenario y:y is extending x:x,
> But it is distinct.
> >and conceptually we assume that they share the text node.
> They do not. In the data model for that fragment, x:x has three
> children: two text nodes of only white-space, and one element node
> being y:y ... the string "text node" is only a child of y:y.
It seems that XML has a lot of different data models or has had a lot
of different data models over the years. That this usage does not
follow any data model for XML doesn't argue against the assumption to
me, since what I'm arguing about is an assumption about common usages
> >so when split they should be
> ><x:x>text node</x:x>
> >and <y:y>text node</y:y>
> I disagree. I don't know of any data model for XML in which text
> nodes are "copied" or considered a property of an ancestor.
> >I think this is a reasonably common usage but I'm not sure if NVDL
> >handles it. I suppose the argument could be made that this is an
> >extremely dangerous and dirty usage, and we would not want to automate
> >that kind of splitting of data.
> On that I agree ... but since the data model does not "split" the
> data as you describe, it fortunately isn't an issue for NVDL.
The argument is not that it should or shouldn't happen, but more
questioning how far extensibility should go and what the needs for
validation of extended documents are. NVDL as a domain specific
language is addressing the use case of validation of extended
documents, as such it has made decisions for handling various
extension scenarios, if there are scenarios it skips then it should be
argued for why they are skipped.
I suggested one reason why one would skip that scenario is that it is dirty.
I think in XML Schema some reasons why common validation scenarios
were skipped was that they did not fit the theoretical model of XML
Schema (this is just an idea on my end), and I think that is a good
argument for a language to make (but if it is so in the context of XML
Schema it just so happened my first ever use of it needed some of the
missing validation possibilities, and this lack has not endeared the
language to me, there can be tradeoffs between purity and necessity.)
A third argument might be that, hey that scenario is hardly ever used.
nobody asks for it.
I think realistically NVDL is something needed for Compound documents
type situations, but there are a lot of things needed in that