In case people are interested: Google has released its Protocol Buffers under the Apache license. It generates (de)serialization code for structured data in Java/C++/Python from a simple schema description.
http://code.google.com/p/protobuf/ Should be pretty simple to wrap the generated code in a Writable. Could be a useful alternative to org.apache.hadoop.record, especially for data you want to use outside of the Hadoop framework. -Stuart
