Since the focus of our discussion seems to have shifted towards patterns, it may be helpful to include some "dark patterns" of the kind that we want to help people avoid. (I'm borrowing the concept from work on interfaces, but would be interested to learn if there is are similar examples from the data object pattern literature).
The pattern that I'm most familiar with at this point are problems with 1:1 Principle violations, i.e. "records" that seem to be about more than one thing. For example:
<metadata>
<title>Mona Lisa</title>
<identifier>http://is.gd/fFbqI</identifier>
<identifier>Inv. 779</identifier>
<format>image/jpeg</format>
<format>oil on board,/format>
</metadata>
From a XML validation perspective, everything here is good. We really don't have a DSP for these kinds of objects, so whether format, etc. can be repeated is unknown, but since we are often dealing with mixed media objects I suspect it would allow repetition. This, of course is one of the dark patterns that encouraged DCMI to create DCAM in the first place. So now, I could re-express the metadata above as not a "record" about one thing, but a set of "descriptions" each about one thing:
<descriptionSet>
<description>
<title>Mona Lisa</title>
<identifier>Inv. 779</identifier>
<format>oil on board,/format>
</description>
<description>
<identifier>http://is.gd/fFbqI</identifier>
<format>image/jpeg</format>
</description>
</descriptionSet>
A DCAM description set is great if you implement it at the start, but it looks like there will be a number of challenges to getting the simple intuitive translation I just did right in an automated way.
In my view, DCAM successfully solved this problem, but it also seems to suffer from some of the problems that RDF has. People seem generally confused about how to use DCAM in practice, especially when the systems that have implemented Dublin Core (i.e. ContentDM) do not allow DCAM-like structures. I am very curious about what will replace simple DC and OAI-PMH and how we can improve our representations of multiple related objects. OAI-ORE is one route (is there guidance for creating OAI resource maps using DCAM?), but there is also a great deal of attention being paid to RDF/Linked Data approaches. RDF/Linked Data, by themselves, don't solve this problem either. I expect we will see lots of confusing RDF like the first example.
Are there other kinds of dark patterns that deserve attention?
Richard J. Urban, Visiting Professor
School of Library and Information Studies
College of Communication and Information
Florida State University
[log in to unmask]
|