Comments regarding Issue #6113 (data model yes/not) and issue #6169 (structured/flat)

Dear all,

as suggested in the last telco, in order to avoid redundancy I've linked the 
main rows of the ontology features discussion table 
(http://www.w3.org/2008/WebVideo/Annotations/wiki/FeaturesTable) to open 
issues (http://www.w3.org/2008/WebVideo/Annotations/issues/open/), mail 
archives (http://lists.w3.org/Archives/Public/public-media-annotation/) and 
telco minutes.  I've also add a new row for the "fragments description" 
discussion, as suggested by Pierre-A.

During the telco the discussion about the need to define a reference 
metadata format remained open. I would suggest to separate this discussion 
in two different topics:

(1) The need to define our own (explicit and formalized) reference metadata 
format: Which I guess is related to open Issue #6113 
http://lists.w3.org/Archives/Public/public-media-annotation/2008Sep/0055.html 
(Requirement "Allowing for a simple API, ...")

(2) Structured/flat: Open Issue #6169 
http://www.w3.org/Bugs/Public/show_bug.cgi?id=6169 (requirements on the 
annotation) and other features (polymorphism or not, segments or not) which 
can apply also to the API even if we discard (1).

My opinion regarding (1) can be summarized as follows:

- The usage of a common API seems to be the cornerstone of our 
interoperability strategy. As an analogy, I like to think in the way the 
Java bytecode provides platform-independence to applications. In our case we 
provide format-independence.
- Typical APIs for accessing data use to be related to a data model (a data 
schema to be stricter), in fact the API and the data schema use to be 
someway isomorphic. In our case, as I understand now, we do not plan to 
declare an explicit data schema (e.g. with XML, RDFS or OWL) to which the 
MAWG API refers. However, IMO, the data schema will be implicit anyway.
- If we have an implicit data model, I do not see any harm in making it 
explicit and formalizing it. But, which would be the benefit?
- Each thing (the API, the associated data schema) have different usages. 
Think for instance in the difference between an XML schema and its 
corresponding JAXB java code.

Concluding: We should select a subset of prioritary use cases (not 
application scenarios). Then we would see if we need an API, a reference 
data schema, or both.

Best regards,

Ruben

Received on Thursday, 27 November 2008 14:28:48 UTC