The inimitable Art Rhyno and I are working on another project together (W-G?), which consists of several parts:

• Refine Art's mind-bending idea of providing a WebDav interface to the OPAC
• Provide an OAI provider for Voyager ILSes (Hello? Endeavor?!)
• Try to put a real and legitimate Z39.50 (and SRW/U) server (that supports exact searches, relations, etc.) on the catalog
  
• Create a new, useful and user-friendly natural language query interface to the catalog.

Art (Mr. Cocoon) is doing the heavy lifting. He's created a webapp that exports the bib database from Voyager to MODS (and downloads it with wget to a mirror). It is apparently easy (for him, not me) to then transform this output to WebDav.

I only really need the first part (although the second part is very "wow-cool"). We'll take the exported MODS records, put OCLC's OAICat upon it for OAI. Then I'll start building a new interface for the OPAC. The databases stay in synch by wget traversing the cocoon output and checking the timestamps of when a record was modified. Very very neat.

We're running into few problems, though.

Sometimes there are typos/data entry errors on records.

This raises an interesting conundrum. There is often the notion that the catalog is the "authority" for the library. Indeed, this is usually true, however it is impossible to completely eliminate human error.

When I was developing course/control, I sometimes had a problem with marcxml records not being valid. Almost every single time something was wrong with the MARC record. Of course, cataloging was always happy to fix it, but this shows chinks in our armor. If our metadata is wrong, who will get it right?

That being said, bad data (in this case) isn't the end of the world. I can live with a margin of error if the interface is even moderately better.