You can now overwrite values returned from .data() with .bind("getData") - returning a value will override any bound value on that element.

Tweaked the .data() event triggering - it now triggers a single 'setData' event, passing in a key value pair of what was changed.

Added $().data(), $().removeData(), and .bind("click!"). .data() and .removeData() handle namespaced data, .data() triggers a "set-KEY" event on all modified elements, and .bind("click!") only triggers a click (and no namespaced events).

De-eval'd selectors and the various DOM methods (will marginally help our speed and make us more compatible with projects like Caja and Adobe AIR). Left a selector eval in for backwards compatibility support of selector plugins.

Fix #2184 by using the jQuery.clean() function instead of a direct innerHTML assignment in the clone() function for IE.

Fixed #2174 by removing the s.dataType == "json" check for possible cross domain since it causes absolute urls to use a <script> tag ajax retrieval even though absolute urls work fine for retrieving local data.  Cross-domain jsonp will still work since the $.ajax() function actually changes the s.dataType to "script" before it reaches this check in cases where it is s.dataType == "jsonp" or s.dataType == "json" and a parameter has =? in it.

Just added support for Accept headers in Ajax requests - defaults to the correct header depending on the type of request that's being performmed. (Bug #1986)

Fixed an edge case in show() where the css says the display should be none.  In that case force 'block' so it will actually show.

Removed check to prevent event from being fixed twice. Unfortunately, in IE this is sometimes necessary with its global event object. Binding both a mousedown and mousemove event is an example.

Fixed a couple problems found in the test suite.  First, IE doesn't like it when text nodes were trying to duplicate their events in clone, so don't do text nodes (that shouldn't have events anyway).  Also the fx module was freezing from a recent update that wasn't quite finished.

Removed a unit test that tested to see if queued objects were of a certain type.  It was decided you can queue anything and it is up to the coder to make sure it was intended.

Fixed #2080 by removing the check for nodeType != 1.  It was put in to limit the queuing to just dom objects (ie not text nodes and comment nodes), but the queuing functionality is being used more broadly than I realized so the check is now removed.