Roadmap for 2.0

overhaul serialize/pretty printing API

#530 XHTML formatting can't be turned off
#415 XML formatting should be no formatting

overhaul and optimize the SAX parsing

see fairy wing throwdown - SAX parsing is wicked slow.

Node should not be Enumerable; and should have a better attributes API

#679 Mixing in Enumerable has some unintended consequences; plus we want to improve the attributes API
Some ideas for a better attributes API?
- (closed) #666
- #765

improve CSS query parsing

#528 support :not() with a nontrivial argument, like :not(div p.c)
#451 chained :not pseudoselectors
better jQuery selector and CSS pseudo-selector support:
- #621
- #342
- #628
- #652
- #688
#394 nth-of-type is wrong, and possibly other selectors as well
#309 incorrect query being executed
#350 :has is wrong?

DocumentFragment

there are a few tickets about searches not working properly if you use or do not use the context node as part of the search.
- #213
- #370
- #454
- #572

Better Syntax for custom XPath function handler

#464

Better Syntax around Node#xpath and NodeSet#xpath

look at those methods, and use of Node#extract_params in Node#{css,search}
- we should standardize on a hash of options for these and other calls
what should NodeSet#xpath return?
- #656
also, clean up or unify the implementations of #xpath-and-friends in Node and NodeSet
- implementations are very similar, but no shared code :(
- decorate nodes in a consistent manner

Encoding

We have a lot of issues open around encoding. How bad are things? Somebody who knows encoding well should head this up.

Extract EncodingReader as a real object that can be injected https://groups.google.com/forum/#!msg/nokogiri-talk/arJeAtMqvkg/tGihB-iBRSAJ

Reader

It's fundamentally broken, in that we can't stop people from crashing their application if they want to use object reference unsafely.