Comments on spec 0.2 (was Re: Introducing the Open Data Definition)

[Danny Ayers]

The general idea and specification seem reasonable, though is similar in so many respects to RDF that good practice would suggest reusing that framework - it's a standard, and there are already stacks of tools built for it. Such reuse could take the form of an expression of the ODD data model in RDF (as a vocabulary) along with a mapping from the ODD syntax to the RDF model. (A convenient approach would be to provide XSLT for ODD to RDF/XML, which could serve as a GRDDL transformation for ODD documents on the Web).

Probably the best intro to RDF:
http://www.rdfabout.com/
See also:
http://blogs.talis.com/nodalities/category/this-weeks-semantic-web

I'd be interested to hear how the format might be used in practice for sync'ing & linking rather than import/export - as in http://en.wikipedia.org/wiki/Linked_Data

Some specific comments -
"ODD t a ke s t he view t h a t existing d a t a p o r t a bility s ta n d a r d s are, d e s pite b eing
p o werf ul, m u c h t oo co m plex for wide s p re a d a d o p tio n."
Which standards? (The nearest to this space would seem to be RDF, and the parts of that would correspond to this specification are no more complex, and RDF has fairly significant adoption).

Dates: RFC 822 is obsolete, superceded by RFC 2822 in 2001, though a better bet for describing data on the Web would be RFC 3339

The UUID
It's not clear why this definition is needed - HTTP scheme URIs can serve the same purpose.

Namespacing
Standard namespacing is covered by the XML Namespaces doc, although allowing the omission of namespaces seems to undermine the whole point in using them in the first place. Similarly the definition around "tags" (closing tags etc) is covered by the XML spec. You could just ref those docs normatively.

What is the namespace for ODD core terms? Proposed media type?

The <remove_entity> component tag
It isn't clear what removal of an entity means. If it means the resource identified no longer has any representations, it would go against the best practice of "Cool URIs Don't Change" (http://www.w3.org/Provider/Style/URI).

I need to think about the approach to metadata association a bit - on the one hand it's a neat way of expressing n-ary relations, on the other it's introducing indirection from the subject of the metadata. Whether that's a good trade-off, I'm not sure.
 
The part about verbs seems to conflate two different things - the process and the data. Also it isn't clear how the definitions for the verb strings would be maintained - "friends" might have different significance across different environments (I'd suggest using URIs here too, along the lines of RDF properties).

Cheers,
Danny.


--
http://dannyayers.com
~
http://blogs.talis.com/nodalities/this_weeks_semantic_web/