XPath-based parser for aggregator

The current parser has lots of issues with parsing different feed formats. The parser of Feeds module was designed a long time ago, when FeedAPI was created, it's not such sleek as well (i remember :) ). Simplepie (http://simplepie.org/) is a huge piece of code. My idea is (at least if we keep aggregator in core :) ), to rewrite aggregator parser based on SimpleXML + XPath.
The basic concept would be the following:

array(
  'title' => array(
    '/xpath/to/title/in/rss',
    '/xpath/to/title/in/atom',
  ),
  'description' => array(
    '/xpath/to/desc/in/rss',
    '/xpath/to/desc/in/atom',
  ),
);

and so on. The hearth of the parser would be a nested array with full of XPathes, maintaining the parser would be like adjusting the elements of the array. And then the parser just has to iterate over the array, trying to fetch the items+properties in order.
There are problems of course: handle namespaces, is simplexml xpath support good enough to do everything?

If aggregator won't be cleaned out, i'm happy to provide patches as well in the near future.