0.4

The :fixup_tags option will try to sort out the hierarchy so elements end up with the right parents.
Elements such as script and style (identified as having CDATA contents) receive a single text node as their children now. Previously, Hpricot was parsing out tags found in scripts.
Better scanning of partially quoted attributes (found by Brent Beardsly on uswebgen.com/)
Better scanning of unquoted attributes — thanks to Aaron Patterson for the test cases!
Some tags were being output in the empty tag style, although browsers hated that. FIXED!
Added Elements#at for finding single elements.
Added Elem::Trav#[] and Elem::Trav#[]= for reading and writing attributes.

0.3

Fixed negative string size error on empty tokens. (news.bbc.co.uk)
Allow the parser to accept just text nodes. (such as: Hpricot.parse(‘TEXT’))
from JQuery to Hpricot::Elements: remove, empty, append, prepend, before, after, wrap, set, html(…), to_html, to_s.
on containers: to_html, replace_child, insert_before, insert_after, innerHTML=.
Hpricot(…) is an alias for parse.
open up all properties to setters, let people do as they may.
use to_html for the full html of a node or set of elements.
doctypes were messed.

Rewrote the HTree parser to be simpler, more adequate for the common man. Will add encoding back in later.

For whatever reason, wrote this HTML parser in C. I guess Ragel is addictive and I want to improve HTree.