Re: Alternatives to containers/collections (was Re: Requirements for a possible "RDF 2.0")

On 20/01/2010 01:33, Harry Halpin wrote:
> On Sat, Jan 16, 2010 at 1:26 PM, Dave Reynolds
> <dave.e.reynolds@googlemail.com>  wrote:

>> A lot of the other issues may well be broken features but they are not a
>> problem.  I.e. they are not what stops people using RDF and when people do
>> use it they work with or around them without massive difficulty.
>
> For those outside the SWIG world, let us note that it is precisely the
> lack of arrays and ordered lists that is leading JSON to be more
> popular in WebApps than RDF, and that the lack of a usable XML syntax
> has led to a number of proprietary
> Atom syntaxes for moving data around, i.e. Microsoft's oData, Google's
> gData, Yahoo's DataRSS.

As I responded to Paul, I can see that the pain of using ordering is a 
significant one.

I do definitely agree with the need to bridge the gap to web developers. 
  I had been seeing this as better done by developing mappings and tools 
and translators on top (e.g. [1]) rather than redesigning the foundation.

> So, yes, sorry folks - RDF 1.0 is not broken, but large amounts of the
> Web find it relatively unusable for applications. So let's fix it. The
> fixes should be relatively simple, and could lead to wide uptake in
> RDF.

It's that "relatively simple" that worries me. You may underestimate the 
difficulty of agreeing consensus standards while preserving appropriate 
levels backward compatibility.

For example, is there already some concrete proposal for handling the 
ordering issue that both you and Paul raise? If so, how does it interact 
with RDFa? Is it backward compatible with existing triple stores? If 
not, what's the migration story? What's the semantics for how it relates 
to collections? How does it affect SPARQL? What's the mapping to RIF's 
list facility? How does it affect OWL 2? etc

My worry is that instead of getting application problems solved and 
useful tools in developers hands, the RDF community disappears into a 
black hole for a couple of years redesigning the standard. That would be 
... unfortunate.

> And more or less every application I've seen uses named graphs
> and agrees with the fixes needed (i.e. putting named graphs into RDF,
> quietly deprecating certain containers and reification) - so let's do
> it.

I have seen several applications that use both containers and 
reification, apparently successfully, and would presumably be 
inconvenienced if they were to disappear. I believe there may be derived 
standards that use them [2], though they may not be active ones. If such 
interests were represented on the working group it may take time to get 
consensus even on those apparently simple issues.

Dave

[1] http://code.google.com/p/linked-data-api/

[2] I seem to recall both UAProf and CIM/XML use Seq though may be wrong 
on that. This impression comes from Jena users referencing those 
standards and I may be confusing correlation with causation.

Received on Wednesday, 20 January 2010 10:05:47 UTC