I really appreciate all of the feedback on this point. Lots of interesting ideas to think about.
Looking at the various responses, it seems to me that by not allowing the repetition of fields, DwC limits its usefulness to information managers such as myself. Some of the work-arounds, such as the GBIF Vernacular Name extension that Peter Desmet pointed me towards, look useful in the particular example that I gave, but won't work in others. It is also a fairly complex process, belying the "simple" part of DwC.
Such a process definitely wouldn't work with some of the other fields that I would like to repeat for our data. I quite dislike the idea of concatenating all the the sample collectors from one specimen into a single field since that will make the process of finding individuals more challenging. It would not be possible to create a relational table for our collectors such as the one for vernacular names.
The other field(s) that I need to repeat pertain to location data. Our dataset currently lists location information in at least five different systems (decimal Lat/Long; deg. min. sec.; UTM; NTS; and verbatim descriptions), and often up to four are used on a given sample. At times the UTM data is generated from degrees Lat/Log, but at other times the reverse is true (and, of course, there is no way of telling from the database alone which is the original). Further, small errors abound in the data that could have crept in during conversions, or possibly even reside with the original data. The data from over 40,000 specimens have already been entered into the database in this manner, and no one is going to go back to double check them all. I desperately want to keep ALL the location data out of fear that we might not present the one accurate measure, and creating a relational table for every geographic point in New Brunswick (let alone the rest of Canada!) is out of the question. (This description of the locational data has been significantly simplified from the actual reality, so please don't start nailing me on technicalities here)
It seems to me, then, that we will have to maintain the data in our own metadata system, and use that to generate DwC (along the lines of what Bob Morris recommended in his first response). That's fine by us, but should, perhaps, be of some concern to a metadata standards working group. Since Darwin Core will not be our de facto standard, its generation and accuracy will be of less relevance to us. I fear the DwC records will become out of date, or start to reflect errors as it maintenance become less important to us. Furthermore, it suggests that we may have to create duplicate sets of data, rather than one set that can be easily harvested for use by other collections.
From my perspective, it would be nice if we could mark this biological data up in one well designed, flexible metadata standard. The Dublin Core group, of course, recognised the importance of flexibility, allowing for Qualified DC along with their simple set, and XML is as popular as it is today because of it extensibility. I would worry that DwC might be painting itself into a corner if it tries to adhere to too narrow a set of rules.
Perplexing stuff, indeed.
Thanks again for all your advice, Geoffrey --------------------------------------------
Geoffrey Allen Digital Projects Librarian Electronic Text Centre Harriet Irving Library University of New Brunswick Fredericton, NB E3B 5H5 Tel: (506) 447-3250 Fax: (506) 453-4595 gsallen@unb.ca