Focus on scraping use cases
- Warn about 
/table/tbody/trin the query, if no results (common mistake)- Consider using HTML5 parser instead
 
 - Shorter method names
 - Guess encoding in the case of bytes input
 - Selector vs SelectorList <- could we have only one class, make it list-like? (like jQuery)
 - Consider adding a shortcut to parse JSON (demjson, to support liberal input JS-like)
 - Add a simple link extraction method
 
Another idea to play with: give a score to a XPath expression representing the probability of it to stop working if HTML markup changes.