resources depending on resources?

[eichin at metacarta.com]

Are resources always "final", or can I construct a resource that
builds upon another one?

The scenario is that I have two layers of "expensive" operations, then
a bunch of cheap ones -- there's "build a local data set from a
source" (web mirroring like I posted before, or nntp download, or SQL
pull).  Then there's "process the local data set to within an inch of
it's life" which is also expensive, but is a single "operator",
ie. the only argument to it is "input data set".  Finally, the tests
are all "poke at that processed data", simple blind-men-and-elephant
observations, but lots of them.

I could just derive the "processed" resource from the "data set"
resource, I suppose, since there's nothing special in the processed
side, and just configure one object there.  It's just that they
already have distinct natural fail-and-cleanup points...