When looking at incidents on disk (as opposed to in the Incident Gatherer's collection), it would be nice to be able to quickly classify them in the same way that the gatherer does.
I'd like two things. The first is a standard for how classification functions should be packaged, to make them into a sort of plugin: a .py file with a well-known name that is called to do classification.
The second is a "flogtool classify-incident" tool, with an option that points to one or more classification plugins. This tool should take the flogfile names in argv and print their classifications, one per line, along with the filename:
% flogtool classify-incident --classifiers ~/tahoe-classifiers.py incident-*
incident-2008-10-07-083746-5nvka2a.flog.bz2 : retrieve-uncoordinated-write-error
incident-2008-10-09-004448-jwynsjy.flog.bz2 : bad-hash-error
Also, the tool should be able to move incident files around based upon these
classifications, so you can move all the well-understood incidents out of the
way and focus on new problems. I'm not sure how to do this while still
preserving the order of the files. There are two separate axes: known/boring
vs unknown/novel, and new/recent versus old.
