It turns out the DVD I have of Cumberland images is incomplete, so I'm waiting for another one by mail, but int he meantime I've decided against doing an automated import; instead, I think it's much safer to provide mapping for the editors from a simple web page which enables them to vet each image before adding it to a record. So I've been refining my Python script, and in the process learning how awful XML/XPath handling in Python is; it's worse than useless, so I'm pre-processing the XML dump of the database to create a much simpler XML lookup file that the Python can use to retrieve the appropriate slugs for each record, based on identifier. I also installed Wand and I'm using that to interface with ImageMagick to create scaled-down versions of the images where necessary (they don't want to upload full-size ones). This is all now working, and I can run it again when I have the DVD to create a page for them to work from.