Lesson 5: Spatial linked data publication is still expensive

In the second phase of the testbed we asked one of the developers to focus on quantification of the cost of publishing spatial linked data.

The main costs are spent converting the various data formats to Linked Data. After the conversion, developers had to go through several iterations of transforming the data to improve its quality.

A second source of significant costs was the absence of tools that allow Linked Data to be exposed to web programmers. There was a strong expectation that those tools would be more readily available, after all, Linked Data is web data. In fact, there were many perfectly valid requests by the web programmers that could not be easily implemented by integrating some existing library.

As the table on page 22-23 of the findings document shows, the costs of setting up a web service for Linked Geodata are still very high.

In practice, some of the costs can be reduced by taking the shortcut of reducing all geometries to 2D points. GeoSPARQL functions between points are mostly supported by existing triple stores and are faster to compute. The downside to this is that much of the geospatial information is lost and very many geospatial queries (‘contains’, ‘intersects’) cannot be performed at all.