Tue Sep 29 09:32:44 CEST 2009

Ramblings File Representation

Instead of building the index page statically, it might be better to
generate it from an enumerator/stream.

Ok, aggregation itself isn't so difficult, the problem is with
propagating cache refresh signals up: the cache mechanism needs a
compose operation.

Hmm..  I suspect this is probably more a problem with my
representation.  It would be nice to figure out a way to detect
structural changes (i.e. add/delete article) as opposed to content
changes, and then simply updating this structure instead of
regenerating it from scratch.

What I really want is each article to be a separate entity.  Can this
be abstracted in that way?

Alternatively, content could be kept completely on-disk, with
expensive operations (like tex->png) cached on top of that, i.e. based
on some hash of the text.

Roadmap: abstract ramblings files as objects with the following
  - header
  - node-list
  - node ref (symbol -> key.value hash)

where data always comes directly from disk, and internal indexing is
regenerated when the file changes.