Digital preservation is not a science, in at least two important ways.
There is no veil
Newton quote, no mystery here.
No experimental evidence is required to see that GeoJSON is a bad format to store CAT scans in. No new theories are required to determine that converting Microsoft Word to PDF may discard a lot of information [SigPropBehold]. These are all constructed systems, built by humans, with documentation to be inspected, whether that is prose or code. To evaluate the format landscape and determine likely preservation options is fundamentally an act of literature.
Emergence
Of course, once the likely options have been determined, based on knowledge and experience, it may well be worth experimenting with different tools and approaches. This can be used to test whether the credible strategies behave as expected, and of course to compare different implementations of the same strategy to see which performs best in a given context.
At this point, when we are attempting to understand the emergent behaviour of complex software systems, then the scientific method can help isolate our investigates from our biases and assumptions.
Criticality of sharing of experimental results in order for this to work (as that’s why it works for science).
Value judgements
Looking back at the early talk of ‘significant properties’, we see ‘essence’ etc.
But Sig in eye of beholder, that identity paper, have shown us this is a lie. The decision to normalise Word documents to PDF is not just a statement about the relative confidence of access to those formats. It is also a judgement about what is important.
This is not a problem. Our decisions, what we choose to preserve, and why, have always involved value judgements. We only want to preserve things that are important to people now or in the future. That’s the whole point.
But, we should be aware that this is a value judgement. We should be wary of taking the technically easier option, without evaluating the damage it may do to the value of the items. Most importantly, the process of judgement itself cannot be made without reference to the needs and values of the community we are preserving these things for.
There is no ‘essence’, beyond all context. Every normalisation pathway, every significant property scheme, contains its own assumptions about where the value lies. This cannot be defined in isolation from use, and we can only plan sensible if we can find ways of engaging with or understanding user needs, and likely future needs.
SO ISSUE BECOMES TYRANNY OF THE MAJORITY, and supporting miniority interests. It becomes political. The information that people maintain interest in will have the most scrutiny and be moved forward. But the niche, the minority, the minor and the overlooked may get left behind.
Conclusions
Scientific, but not a science. Evidence-based policy making. Stakeholder engagement. Closer to politics than physics.
