Note that this is a pretty straightforward patch where the bulk of the pep8 parser code is copy-pasted directly from the pylint parser. A better way to do this would be to implement an abstract regex-based parser which both the pylint and the pep8 parsers would then extend, defining their particular regexes, the mapping of regex group indexes to violation fields and the mapping of message types to violation severity values.
I am willing to implement this improvement if the maintainer(s) can give me an idea as to where this abstract parser superclass should be placed (it looks like hudson.plugins.violations.parse would be the package for this but I'd like a green light nonetheless).
Added patch as SVN diff