Jeni's Musings

Hi Jeni,

I think that base of misunderstanding between NVDL and XML Schema folks is really very simple — one like loosely coupled and others tightly coupled systems and approaches. As you say, if you use xsi:schemaLocation with multiple schemas with wildcards, you can validate compound documents. But this means that authors of original schemas had your usage scenario in mind, that you believe that after 20 years you will be still using W3C XML Schema for validation (and thus you will don’t mind xsi:schemaLocation in your instances) and that you are always validating your document again one fixed set of schemas.

Of course I (and many other people involved in NVDL) do not believe to this. In general, schema should not be specified in instance, because over the time validation technologies are evolving, and you probably do not want to edit your document just to point to a new and better schema. Moreover I quite often find useful to validate document against different schemas to check for different constraints for different purposes. And finally, with NVDL you can create compound documents utilizing XML vocabularies which were not originally developed for such purposes (or at least their schemas were not written witch such usage scenario in mind). IMHO this is very pragmatic approach as almost noone can imagine and foresee all possible usages of schema his/her developed.

Sorry for short blurb, but alpha version of web front-end for JNVDL was just launched at http://relaxed.vse.cz/nextgeneration/

Regarding XProc: You can certainly simulate NVDL behaviour with XProc, but I suppose that in almost all cases XProc code will be much more longer and procedural then corresponding NVDL very declarative code. But this is the case for specialized languages like NVDL.

But I think that XProc has much greater overlap with DSDL Part 10 (Validation management), then with NVDL. It can even turn out that XProc is sufficient for Part 10 and will be just adopted for it.

I’m thinking about “CreoNVDL” — something like NVDL operating on overlapping markup. NVDL approach could be very useful for orthogonal vocabularies — like XHTML/DocBook/SomeOfficeXML + revision tracking + comments. I think that for certain usecases this could be more effective than somewhat complex content models which you can get with Creole (or at least complex for someone who is not yet very used to “overlapping” idea).

Anyway, I was glad that I can met you personally during XTech.

Jirka