Today I was looking for the best way to write the feed reader and these are my thoughts:

* XmlTextReader seems to be the fastest xml reader and looks nice but I don’t like the pull model very much and since the library isn’t mission critical I’ll consider to use it only if I don’t get anything else working (maybe I could try to implement more than one reader… I don’t know)
* XmlDocument is handier than XmlTextReader cause uses the DOM but it’s slow (not very much to being honest, I can’t imagine a 5+ MBytes Atom feed) and I don’t need editing or inserting any data at all.
* XPathDocument + XPathNavigator is quite reasonably the right direction to go. Fast, readonly and exposes to me all the power of XPath.

So, after re-seeing MSDN-TV’s Don Box episode called “Passing XML Data Inside the CLR” I decided to give XPath a try.

Stay tuned.