- From: Henri Sivonen <hsivonen@iki.fi>
- Date: Mon, 17 Nov 2008 18:18:02 +0200
- To: Boris Zbarsky <bzbarsky@mit.edu>
- Cc: elharo@metalab.unc.edu, public-html <public-html@w3.org>, www-tag@w3.org
On Nov 17, 2008, at 17:50, Boris Zbarsky wrote: > I think it is (simply by virtue of being a state machine; it should > be possible to automatically verify that all transitions for all > inputs in all states are defined; once that's done, the spec is > certainly unambiguous). Each state in both the tokenizer and all but three states of the the tree builder have an "Anything else" entry, and the three tree builder states are otherwise exhaustive even though the words "Anything else" don't appear. As far as I can tell, the major source of ambiguity is the same one that causes a huge hole in XML's definition of well-formedness: If the input byte stream has an unusual character encoding, agents that support the encoding and agents that don't arrive at different DOMs. Plugging this hole would require specifying a fixed finite set of encodings that agents MUST support and specifying that agents MUST NOT support encoding from outside the set. -- Henri Sivonen hsivonen@iki.fi http://hsivonen.iki.fi/
Received on Monday, 17 November 2008 16:18:52 UTC