Trustworthy Preservation Planning

The mission of digital preservation is to overcome the obsolescence threats that digital material is facing on the bitstream, the logical, and the semantic level, and to provide continued, authentic long-term storage and access to digital objects in a usable form for a specific user community. This requires preservation actions to be carried out when the original environment of digital objects is unavailable, to either recreate it (emulation) or transform the objects’ representation into a form usable in a new enviroment (migration). A variety of preservation actions exist. Quality varies across tools; properties vary across content; usage and requirements vary across users and scenarios; risk tolerances, preferences, costs, and constraints vary across collections, organisations, and environments. Finally, all of these factors are subject to constant shifts that have to be detected and handled. The mission of preservation planning is to ensure authentic future access for a specific set of objects by defining the actions needed to preserve it. The core problem of preservation planning is a domain-specific instance of component selection and can be correspondingly reformulated and modelled. The arising research questions are threefold: (1) How can we select the optimal preservation action for a given setting? (2) How can we ensure trustworthy preservation planning? (3) How can we ensure that decision processes scale up? This thesis describes a systematic framework for preservation planning, comprising a multi-objective decision making method, workflow and tool for creating preservation plans for sets of digital objects. Policies as high-level influence factors model environmental constraints and specify organisational preferences. Preservation planners empirically evaluate potential action components by applying automated measurements in a controlled environment and select the component that is optimal with respect to the particular requirements of a given setting. We present a distributed architecture for preservation planning which integrates planning, actions, and characterisation, with the planning tool Plato at its core. The tool implements the planning method and creates solid, well-documented preservation plans. It has experienced significant uptake in the digital preservation community. We describe examples applying the approach to a number of real-world business decisions, discuss limitations of the approach, and identify a key challenge of evaluation. We further analyse the influence factors to be captured and evaluated and categorise them in a taxonomy. We show that a majority of the criteria can be evaluated by applying automated measurements under realistic conditions, and demonstrate that controlled experimentation and automated measurements can be used to substantially improve repeatability of decisions. This reduces the effort needed to evaluate components, enables scalability, and supports trust in the decisions because extensive evidence is produced in a repeatable and reproducible way and documented along with the decision in a standardised and comparable form.