Issue 947: HTML emitter vs. getAttribute() inconsistent

Fix some inconsistencies in HtmlSanitizer and HtmlValidator.
This makes HtmlValidator more paranoid, and fixes the HTML attribute
schema to whitelist the accesskey attribute.

I tried repeating the issue in a number of revisions from the last 6
months, but couldn't.  But the schema fix is necessary, and the sanity
check can't hurt.


Submitted @3291

[Jasvir, 2009-02-21 01:25:26]

http://codereview.appspot.com/21046/diff/1/6
File src/com/google/caja/plugin/HtmlSanitizer.java (right):

http://codereview.appspot.com/21046/diff/1/6#newcode239
Line 239: private boolean removeBadAttribute(DomTree.Tag el, Name attrName) {
Why is it ok to remove the (el == null) check here.  Isn't tag == null here if
!(htmlRoot.parent != null && htmlRoot.parent.node instanceof
MutableParseTreeNode)?

[MikeSamuel, 2009-02-21 01:37:47]

http://codereview.appspot.com/21046/diff/1/6
File src/com/google/caja/plugin/HtmlSanitizer.java (right):

http://codereview.appspot.com/21046/diff/1/6#newcode239
Line 239: private boolean removeBadAttribute(DomTree.Tag el, Name attrName) {
If we can't remove a bad attribute, I would rather the whole thing fail
catastrophically rather than possibly return an unsafe output.

[Jasvir, 2009-02-25 03:23:22]

LGTM