[ckan-dev] RDF Serialization?

[Rufus Pollock]

On 20 June 2011 16:14, Friedrich Lindenberg
<friedrich.lindenberg at okfn.org> wrote:
> Hi Kendra,
>
> On Mon, Jun 20, 2011 at 4:03 PM, Kendra K. Levine
> <kendra.k.levine at gmail.com> wrote:
>> One question that I still have is about RDF serialization. I've heard
>> conflicting stories about it in CKAN, and I'm not sure if there's any
>> consensus on the issue. I just sort of assumed we would be using RDF, but I
>> would like get a better sense from the community.
>
> there are at least 3 different ways to do it, which I'd call the "old
> will waites way" (old www, it has to be given a name), the "new will
> waites way" and what we did for publicdata.eu:
>
> * old www: Python batch job generates plain files to be stored on
> disk, these used to be served from semantic.ckan.net.
> * new www: Go daemon traverses several catalogues, keeps them in
> memory as an RDF graph and serves them out to the RDF hungry-masses.
> * pdeu: We're mounting an RDF representation of packages at
> /rdf/package/NAME and writing to a 4store whenever things change in
> the RDBMS. The store SPARQL is then reverse-proxied to /rdf/sparql.

IMO key point is that going forward: we will generate an RDF
representation as part of core CKAN (rather than having to depend on
an external service). This does not mean external service may not also
be really useful but we will have at least a basic mapping in core
CKAN. This means something like option 3.

Having SPARQL and 4store would not need to be part of this but would
be a nice extra (and is orthogonal to how we do the mapping).

@Kendra: what exactly are the conflicting stories you mention?

Rufus

[...]