Last week, the Second Linked Data Meetup London was held at the University of London Union. There were several compelling presentations discussed on Twitter including the BBC's use of Linked Data for their Wildlife Finder app. One of the many promising topics to emerge from the day was the introduction of a new Linked Data API. While there have been other Linked Data APIs (Pubby and irON), this API has the more narrow goal of lowering the bar for non-SemWeb developers to access these rich data collections. It is intended as a simple RESTful layer that returns JSON representations of RDF collections backed by a SPARQL endpoint. The API was primarily developed by Dave Reynolds, Jeni Tennison and Leigh Dodds. At the "How the Web of Data Will Be Won" talk by Jeni Tennison and John Sheridan, they emphasized that extending their successes on exposing government data in a linked fashion will require a focus on usability to attract new developers.
The specification of the new read-only API details the design goals and decisions. Configuration files will coordinate the mapping of URIs to endpoints through templates and binding variables. Users will ultimately be able to negotiate the results into a variety of forms (JSON, Turtle, RDF/XML) or apply XSLT transformations to produce information in the form they want. Query parameters will provide a capacity to shape the results based on type, size, etc. Some examples include:
/doc/school (return all schools from a government site)
/doc/school/12345 (return information about a specific school)
/doc/school?type=primary (return all primary schools)
/doc/school?min-size=300 (return all schools with more than 300 students)
/doc/school?near-lat=0&near-long=0&distance=5 (return all schools within a specified distance from a location)
These APIs are clearly more accessible to developers, but they are intended to be a graceful on-ramp, not a replacement for the richer semantic queries and data sets. The SPARQL queries, endpoints and RDF will still be accessible directly or through a variety of transformations. The larger point from Jeni and John was that efforts like http://data.gov.uk to make government information accessible remains difficult. Serious consideration must be given to long-lived URIs, versioning, integration points, etc. While trained data experts wrestle with those issues, the raw data can still be used by more conventional Web developers, so it is time to encourage them to do so.