One part of the solution: Make it possible for people who are building archives of filk music to let others know what they have. There is legislation under consideration to make it easier to restore "orphan works" -- works with no known claimant to copyright -- to publication. Just the fact of knowing that someone has a copy of a song that might be listenable by private arrangement may be valuable for music historians, or for the heirs of the writers and performers. We're talking about long-term considerations here.
The heart of the Filkive project is a metadata model. There are a variety of existing data models -- Dublin Core, METS, MODS, EAD, etc. -- but none are quite right for this project. A major consideration here is to keep the data lightweight, yet complete enough to be useful. This requires some specialization. Fortunately, it's easy to turn one object model into another, using XSL transformation tools, and we hope to provide translations to and from standard metadata schemas as part of this project.
The original description of the Filkive project is available here.
Current pieces of the project, more or less functional, include a converter from tab-delimited text (.tsv) to XML and a simple data entry application.
All the code is in Java, so Filkive can be ported to all major operating systems that provide Java 1.4 support.