Wed Apr 1 11:17:52 CEST 2009

implementing aggregation

  - update: currently when file changes it gets recomputed
  - mutation is used to index the list of articles.

  - separate indexing (the node datastructure) from a pure
    sequential list of articles

OK.. Using the straightforward approach creates problems.  I'm not
sure which evaluations get triggered again and again..

TODO: eliminate whole-datastructure traversals.