Re: GoodRelations Light

Hi Bob:
On May 3, 2011, at 12:26 PM, Bob Ferris wrote:

> Hi Martin,
> On 5/3/2011 11:49 AM, Martin Hepp wrote:
>> Hi Bob:
>> gr:name and gr:description are Google-specific. If you use dc:title / dc:description or rdfs:label / rdfs:comment, your data may not be included in Google Rich Snippets.
> I really don't understand such a design decision by Google. This is contra-productive.
It was my design decision, eventually, and in fact I see arguments in its favor.
>> You can easily use multiple properties for the same literals, of course or define a SPARQL CONSTRUCT rule or a mapping via owl:subPropertyOf.
>> The motivation for gr:name and gr:description was that Google prefers a minimal number of namespace declarations, and foaf:page and foaf:depiction were more important for me at that point, so I sacrificed on the textual properties. Also, for E-Commerce applications, it may make sense to have a specific GoodRelations property for textual elements.
> Yes, I guess, it will still take some time that traditional search engine developers will hopefully grasp the advantages of multiple namespaces. This is one reason that makes the difference between database schemata (or similar approaches where everything is redefined everytime) and a kind of "shared understanding" which can be created by the usage of common (terms of) ontologies, such as DCTerms.
> We should force the application for multiple namespaces (vocabularies), instead of avoiding it.

Conceptually, multiple namespaces are good.
For a markup-centric ecosystem, multiple namespaces are burdensome.
The same applies to URIs for codes vs. authoritative literals.

For impact, you have to balance out the pros and cons of those, and the academic linked data community seems to focus on conceptual beauty alone.

A typical GoodRelations recipe requires about ten lines of RDFa markup. In many RDFa patterns, the namespace declarations span 5 - 10 lines. That is confusing for many developers.

In this direction, the idea of a "pedantic Web" movement is a contradiction in itself, because a key characteristic of the Web is that is not built on pedantry, as the Web as a whole does not break if there is a lot of buggy markup and incomplete understanding of specifications by the majority of developers.

When I started my first Web page, all I had to grasp was

- some basic HTML patterns, copied from a the source code of a friend's page and
- the basic usage of an FTP client.

When someone wants to join the Linked Data Movement now, we expect him or her to

- understand the RDF data model
- read RFC 2616
- understand the Linked Data Publication Principles
- understand various syntactical forms
- understand the interplay between RDF and RDFS/OWL
- understand SPARQL
- understand lots of different vocabularies

That's too much. It's okay to expect this from a developer team in a start-up building an *application* on top of that data, but not from the millions of past-time Web developers. 

> Cheers,
> Bob

Received on Tuesday, 3 May 2011 10:49:01 UTC