Progress so far:

• Implemented default responses for all six verbs, mostly with default info and placeholders.
• Fully implemented the ListSets response, which has quite a range of sets.
• Implemented the core of GetRecord, which calls out to a getDocRecord() function that does the real work.
• Implemented about half of the getDocRecord() function; it can return a header (fulfilling the needs of ListRecords), and I'm just getting started on the Dublin Core metadata output in the <metadata> tag.
• Fully implemented ListMetadataFormats (easy, because we're only supporting oai_dc).
• Fully implemented Identify.

One thing I'm currently undecided on is whether I should bother with the resumptionToken functionality. If I do, that means I'll have to cache the parameters of the request in the db somewhere and retrieve them in response to the token, which is a bit of a pain; I'm more inclined to let the whole thing run, and only worry about the resumption token if it seems likely that the results will be too large to handle.