It might be most fruitful to organize the papers into yearly files:

• 1989.scm
• 1990.scm
• 1991.scm
• ...

Since the set of papers published in a particular year is fixed, it's easier to keep track of progress. When the papers are organized by topic (as they are now), each file can grow without bound and there are some ambiguous cases where a paper concerns more than one topic.

The organization by topic is very useful for browsing; we can keep it by adding tags to each paper. Then the topics can be specified as set operations on tags.

@amirouche What do you think about this approach?