In message <Pine.LNX.4.33.0102122007010.4389-100000@jack.pacbell.net>, Eugene E
ric Kim writes:
>An XML property set has been defined, but that's not the big picture.
>
>The notion of groves came out of the SGML world, but is not limited to
>SGML or any markup language for that matter. In order to link to a
>chunk of data -- be it a sentence in a Word document or a still frame in a
>movie -- you need to be able to address that data. In order to address
>that data, you need to know the document's data model.
>
>What you don't want to do is force all document types into a single data
>model. That's ludicrous. Presently, there isn't even one standard data
>model for XML documents.
Of course. But DTDs and XML Schema definitions (and Relax
definitions) are all syntactically-defined data modelling
languages for XML. What does a grove give beyond that?
At this point, the building blocks I'm using for modelling the
repository structure are simple: StructureNode (ordered list of
children), TextNode and DataNode. My plan has evolved to using these
building blocks to move forward to general document classes (e.g. XML,
MP3 audio, MPEG/2 video, etc.) with plugins and then specific data
models within these classes through DTD/Schema definitions.
The repository becomes a live, peer-to-peer document store with an
extensible system (plugin + schema) for moving from abstract to
specific structure and then these schemas become a bridge to the
semantic layer.
The classes of repository interfaces are:
1 Structured document storage/manipulation
2 User and permission mgmt.
3 Linkbase management
4 Content-based search
5 etc.
>Ideally, what you'd like to have is a metalanguage for describing a
>document's data model. That way, all you need to address a document (and
>subsequently link to it or do view control or whatever) is the document, a
>description of the document's data model, and an engine that can read the
>data model description and parse the document accordingly. Note that this
>also allows you to define multiple data models for a document type, so
>there doesn't have to be only one way to address a document.
I gues my basic question is "how does this differ from XML schema
mechanisms?"
-------------------------------------------------------------------------------
Lee Iverson SRI International
leei@ai.sri.com 333 Ravenswood Ave., Menlo Park CA 94025
http://www.ai.sri.com/~leei/ (650) 859-3307
This archive was generated by hypermail 2.0.0 : Tue Aug 21 2001 - 17:58:01 PDT