Re: Metadata access

On Apr 26, 2011, at 4:30 , Silvia Pfeiffer wrote:

> 2011/4/26 Joakim Söderberg <joakim.soderberg@ericsson.com>:
>> Dear all,
>> Recently the Ontology for Media Resources and API has been discussed in your forum. On behalf of the Media Annotations WG, I would like to share our response below.
>> 
>> We are following the discussion on access to media metadata with high interest, as this problem is central to the scope of the Media Annotations WG. From your discussion, it seems that you are aiming for a simple and flexible solution. This is also what MAWG has in mind, and thus we would like to understand better which of your requirements are not satisfactorily addressed by the MAWG specs.
>> 
>> First of all, as some of you probably know, MAWG is working on two specs: an Ontology for Media Resources and an API for Media Resources. The ontology is defined informally as a set of basic properties of media items, together with mapping to a number of standards and formats. We are convinced that this approach (what Danny called "upper ontology") is necessary to establish interoperability, i.e., enable the use of the metadata without knowing about the specifics of the source format. Note that the metadata properties defined by MAWG are not at all a required set of properties to be provided - depending on the source format only few of them may be available. In addition, the ontology is defined formally as an RDF ontology. The ontology can thus be used without the API, e.g., included in HTML using RDFa.
>> 
>> The API builds on top on the ontology, providing a way of querying the set of properties in a homogeneous way independent of the source formats. It seems that most of the criticism in the your discussion actually targets the API, i.e., the fact that getMediaProperty is called with a set of parameters and the complexity of the return structure. Actually, our first draft was more like the .language example in Danny's post, but we got the feedback that (i) browser manufacturers would prefer an API with as few functions as possible and (ii) web developers would rather get a result list and iterate through the results of the different properties than calling distinct functions for each of them (Silvia attended this discussion back at TPAC 2009).
> 
> 
> Indeed. And my continuous feedback was that IMHO a name-value-pair
> interface would be the most complexity that would be required. I don't
> like the complexity of the return structures - never have.

would it hurt to define both APIs?  After all, the simple one is trivially (if not optimally) implemented on top of the 'complex' one...

David Singer
Multimedia and Software Standards, Apple Inc.

Received on Tuesday, 26 April 2011 22:53:08 UTC