People, organisations, and machines around the globe make use of web technologies to communicate. For instance, 4.16 bn people with access to the internet made 4.6 bn pages on the web accessible using the transfer protocol HTTP, organisations such as Amazon built ecosystems around the HTTP-based access to their businesses under the headline RESTful APIs, and the Linking Open Data movement has put billions of facts on the web available in the data model RDF via HTTP. Moreover, under the headline Web of Things, people use RDF and HTTP to access sensors and actuators on the Internet of Things.
The necessary communication requires interoperable systems at a truly global scale, for which web technologies provide the necessary standards regarding the transfer and the representation of data: the HTTP protocol specifies how to transfer messages, besides defining the semantics of sending/receiving different types of messages, and the RDF family of languages specifies how to represent the data in the messages, besides providing means to elaborate the semantics of the data in the messages. The combination of HTTP and RDF -together with the shared assumption of HTTP and RDF to use URIs as identifiers- is called Linked Data.
While the representation of static data in the context of Linked Data has been formally grounded in mathematical logic, a formal treatment of dynamics and behaviour on Linked Data is largely missing. We regard behaviour in this context as the way in which a system (e.g. a user agent or server) works, and this behaviour manifests itself in dynamic data. Using a formal treatment of behaviour on Linked Data, we could specify applications that use or provide Linked Data in a way that allows for formal analysis (e.g. expressivity, validation, verification). Using an experimental treatment of behaviour, or a treatment of the behaviour's manifestation in dynamic data, we could better design the handling of Linked Data in applications.
Hence, in this thesis, we investigate the notion of behaviour in the context of Linked Data. Specifically, we investigate the research question of how to capture the dynamics of Linked Data to inform the design of applications. The first contribution is a corpus that we built and analysed to monitor dynamic Linked Data on the web to study the update behaviour. We provide an extensive analysis to set up a long-term study of the dynamics of Linked Data on the web. We analyse data from the long-term study for dynamics on the level of accessing changing documents and on the level of changes within the documents. The second contribution is a model of computation for Linked Data that allows for expressing executable specifications of application behaviour. We provide a mapping from the conceptual foundations of the standards around Linked Data to Abstract State Machines, a Turing-complete model of computation rooted in mathematical logic. The third contribution is a workflow ontology and corresponding operational semantics to specify applications that execute and monitor behaviour in the context of Linked Data. Our approach allows for monitoring and executing behaviour specified in workflow models and respects the assumptions of the standards and practices around Linked Data. We evaluate our findings using the experimental corpus of dynamic Linked Data on the web and a synthetic benchmark from the Internet of Things, specifically the domain of building automation.