Whether you run a website or a business, it is easy to think of all of your data as an asset—as something that produces business value and therefore should be guarded like any intellectual property. However, if you try to carefully differentiate your company’s value from your competitor’s, it often becomes apparent that there is only a small amount of data that is truly valuable and that provides your competitive advantage. The vast majority of data is often simply the glue that connects your valuable data to the rest of the world and makes it possible to use.

This distinction can be described as follows. The things that produce business value and allow a business to differentiate are what Geoffrey Moore, a technology business analyst, calls “core.” Everything else that goes on in a business that is not a part of value differentiation is deemed “context.” Moore points out that to optimize business value, organizations should focus on those things that are core and treat everything that is context as a commodity service, outsourcing to lower-cost providers whenever possible.

From this perspective, the data from which your company derives value—the information about your users, the data produced through your R&D, the data that your competitors don’t have access to—becomes core. All the other data that is used to make sense of the core data—geographic metadata used to organize your sales force, lists of industry standards that you need to comply with, databases of vendors that you keep on hand for convenience, product databases that everyone in your market must maintain—all of this is really context. Money and time spent maintaining and growing core data helps grow the company. On the other hand, resources spent on maintaining context data are overhead and operating costs—potentially a liability, and best if minimized.

By making data available to the Giant Global Graph, organizations have an opportunity to share the burden of maintaining context data by creating a commons of data. Shared data is a public good that lowers everyone’s overall cost of doing business. And unlike other types of public goods, there can be no tragedy of the commons with information because information it can never be used up. If one group takes more than they contribute, the overall community is not hurt. There is only upside as people contribute into the commons.

Public data is only useful if you understand how you are allowed use it. Copyright is a complicated area of law, and it is often unclear if there are limitations on the use of public data. Traditionally, the metadata about permitted uses of information has been provided in legal documents intended for humans—and even then, comprehensible by only a few of us. For a truly global graph to work, machines must not only be able to discover data and join it to existing information, but they must also be able to understand how the data can be used.

Creative Commons is a nonprofit organization that provides standardized ways to grant copyright permission to creative works. To convey the rights granted by information providers to data consumers in a machine-readable way, Creative Commons has developed the Rights Expression Language (ccREL), an RDF vocabulary for describing licenses. The vocabulary can be broken into two sets of predicates: Work Properties, which attach license statements to creative works, and License Properties, which describe the legal aspects of a license. See Figure 11-3.

Figure 11-3. Creative Commons licenses

Creative Commons provides an array of pre-constructed licenses, which make use of the License Properties to describe how data can be used. To use Creative Commons licenses, you utilize the Work Properties to specify which resources are associated with which license. For instance, to grant rights so that anyone can use the MovieRDF dataset as long as they give attribution to Freebase, the source of the data, we make use of the XHTML license predicate to connect the resource for the dataset to the Creative Commons Attribution license. We then use Work Properties to indicate the name and URL that should be used for making proper attribution to the data creator. In N3, these RDF statements would be:

@prefix cc: <http://creativecommons.org/ns#>
@prefix xhtml: <http://www.w3.org/1999/xhtml/vocab#> 

<http://semprog.com/psw/chapter4/movierdf.xml>
      xhtml:license <http://creativecommons.org/licenses/by/3.0/>
      cc:attributionName "Freebase - The World's database" ; 
      cc:attributionURL <http://www.freebase.com/> ;

You can find additional predicates for specifying other information about the licensed work and the full set of License Properties for constructing new licenses at http://wiki.creativecommons.org/CcREL.

When you choose a license, think carefully about the restrictions that you place on your data. While selecting a license that disallows commercial usage or derivative work may seem like a conservative and prudent business decision, this often results in a lack of adoption of the data that you are trying to share. It can even lead to competing efforts to open up the same data with fewer restrictions, and can result in a community backlash against your good intentions.

If you’re trying to engage with a community of users and businesses, the Creative Commons Attribution license (CC-BY) is a liberal license that allows users and businesses to use your data as long as they credit you when and where it is used. It allows your users and partners to freely build on your data, while ensuring that nobody claims that your data is their sole property.

When thinking about semantic data ecosystems, it is tempting to partition the world into data providers and data consumers. At first glance, it seems like these types of distinctions could help identify the technologies used by participants in different parts of the data ecosystem. Consumers of web pages use web browsers, while producers of web content use content management systems, application servers, and web servers. Hopefully we have demonstrated a symmetry in the consumption and production of semantic data. A consuming application will typically have a local triplestore in which to join semantic data from multiple sources, and will use graph queries to extract the information used by the application. A producing application will use a local triplestore on which to run graph queries, which are serialized out for specific requests. Both types of participants in the system use the same machinery, and both can consume and produce data equally easily.

For this reason, networks of semantic data are more than point-to-point exchanges of information. Semantic applications are about integrating multiple sources of data through easy, standardized patterns. A good consuming application can reveal things that no singular source of its data “knew.” The application can then republish that information in a machine-readable form for consumption by other applications.

This approach, where consumers are also producers, forms a data cycle. Information is transformed, augmented, and refined by a loose coupling of services that can be joined end-to-end by other applications. Each service transmits data in a standardized way, packaging metadata with data for use by other downstream services. Through the use of well-known vocabularies and license attribution, services further down the chain may have no contact with services earlier in the chain, but will know precisely how the data can be used. This vision of machines helping one another to help humans has long been a vision in science fiction, but with growth in the Linked Open Data community and applications that implement the trivial “serving” functions we have demonstrated, we are on the cusp of making this a reality.