I am reviving quite an old thread, but perhaps this will become useful to someone:
Recently I became interested in the so called Semantic Web, which led me to Redland library, one component of which is Raptor
parser, which claims to parse "RSS tag soup including all versions of RSS, Atom 1.0 and 0.3". Of course, it parses them into RDF, but it is really well suited for that sort of thing (although I suppose using Redland just for feed parsing is a bit of an overkill...).
I have written basic bindings for Redland: cffi-redland
. I have no actual project to use it for, so I have no motivation to exhaustively test it at this time (especially string memory ownership worries me), but if someone was to use them I could probably work on those a bit more.