Lists Home |
Date Index |
On Monday 28 January 2002 09:44 am, Bullard, Claude L (Len) wrote:
> I think what Steve said was about right: a light single
> point of contact. But, with namespaces, we hide a property
> in the code and expose it in the data only if we want to
> (really, in the text). We specify it in the instance. A
> #FIXED attribute specifies it in the DTD or schema and only
> exposes it in the instance if we want to, or because, the
> DTD isn't processed reliably. Well-formed processing builds
> in unreliability just as the use of internal subsets builds
> in unpredictability.
True enough.... though I think this boils down to an issue of packaging and specification. The "standalone" hack in XML is, in some ways, a step in the right direction, in that it tells the system that we need a bit more than just this instance.
A good packaging mechanism (meaning more than just the physical packaging) such as XAR+ would allow the person producing the document to put all the needed bits and pieces together.
I'm trying to remember when people started assuming that a naked instance was in and of itself useful.... when the XML activity was kicked off, the charter explicitly included bits that would make XML actually useful as a generic markup language (stylesheets and linking in particular). Somewhere along the line, we lost focus of the need for things to *process* XML as part of the application.