[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] new version of graph model paper posted

On Thu, 2002-04-11 at 00:02, Eugene Eric Kim wrote:
> The term that I settled on was "hyperdocument."  (Boy, that one was just
> staring at me in the face. :-)  I think the flaw with Groves and NODAL as 
> a candidate data model for the OHS is that they are too document-centric 
> -- NODAL less so than Groves.  They express a data model for a document, 
> but depend on something external to relate data between documents.
> The point I try to get across in my paper is that there is no need for 
> such a distinction, and in fact, there is much to gain by not separating 
> the two.  A hyperdocument, as I express it, expresses data and their 
> relationships with each other.  Documents are simply two-dimensional views 
> of a hyperdocument.    (01)

Well, I'm kind of with Murray on this one.  I'm glad you've had your
insight and seen the light on documents as views into the underlying
data representation, but for me (and clearly some others), this is not
a new insight.  NODAL is somewhat document-centric in that the type
system is built around document schemas, but the actual data repository
is only really loosely connected to the documents involved.  There is a
"natural" and perhaps assumed mapping between a set of nodes and their
"home" document, but these nodes can be arbitrarily reused in other
documents and mixed and matched with other "compatible" types.    (02)

The real difficulty in managing the relationship between a granular,
reusable data repository and a set of encoded documents is expressing
the subset of nodes that will form the document.  If there is some
sort of DAG or tree descended from a root node, then this is easy and
implicit.  It is the natural thought from someone with a passing
familiarity with the DOM and even with the NODAL as defined in the
White Paper, but it is certainly not the only possibility.  Ideally,
we would be able to specify a nodeset as the result of some query and
then use some document encoder to translate that result into a new
document.  This would be how Doug's views would be expressed in NODAL.
What is the query language though?  How do the encoders get expressed? 
There's some definite machinery to get from here to there.  But the
the thoughts are relatively familiar.    (03)

Lee Iverson                    
leei@ece.ubc.ca                Dept of ECE, 2356 Main Mall
http://www.ai.sri.com/~leei/   Vancouver BC Canada V6T 1Z4
Office: (604) 822-3381    (04)