If rawdog crashes while updating a feed, it shouldn't forget the feeds it's
already updated. Perhaps have an exception handler that keeps a safety copy of
the state file and saves where it's got to so far?

Improve efficiency -- memoise stuff before comparing articles.

Detect duplicate articles (BBC News).

Detect edited articles by guid.

__comments__ from the feed (and anything else worth having?).

Option to choose whether full content or summary is preferred.

Review expiry logic: is maxage=0 the same as currentonly?

Escape characters in text/plain?

Plugin that gets run on new articles.

OPML listing -- needs feed type.
Do for now as separate program once config parser separate.

Error reporting as feed (or as separate display?) (Dorward)

RSS output. raymond@dotsphinx.com

Feed schedule spreading.

Timeouts should be ignored unless it's been getting timeouts for more than a
configurable amount of time.

Replace feed_hash with feed_id; let the user specify the ID in the config file
(defaulting to the hash); let the user use the ID instead of the URL when
specifying a feed.

Add a needs_update() method to Feed; make Rawdog call that on all the feeds
(when not being forced) and then call update() on each of them that needs it.

Make feedparser take a urllib2.Opener rather than proxies etc.

Where articles can be shown to have been updated (by unique ID), only show the
latest version.

For next release:
- generalise templating (so the show template option takes a name, etc.)
- require Python 2.3; kill 2.2 urllib2 workaround
- use unicode.encode('ascii','xmlcharrefreplace') if possible?

Abandoned:
- __feednumber_mod_N__ (for easier default styles, i.e. _mod_4 counts 0-3,0-3)
  (didn't really work for lots of feeds)