RE: MathML-in-HTML5

On Tue, 3 Oct 2006, Paul Topping wrote:
> If the XML island is not well-formed (I hope I'm using the term 
> correctly), then the Microsoft HTML parser is likely to get confused. 
> For example, if a MathML fragment consisted of unbalanced tags, or was 
> missing its </math> ending tag, it may tell MathPlayer to render the 
> rest of the document, though they probably have heuristic rules to 
> prevent that in most cases. Bottom line is that there's no validation 
> going on.

If there's no syntax checking, and the error handling rules aren't 
well defined, then it's just tag soup.

> Just to confirm, David is correct, MathPlayer treatment of bad input 
> (ie, malformed MathML) by putting it in a red box is its own invention. 
> IE plays no role. MathPlayer is just given a DOM tree rooted at the 
> <math> node and it can do as it pleases. (Actually, MathPlayer has 
> access to the entire DOM and more if it wants it.) One consequence of 
> this is that (a) MathPlayer doesn't have to implement an XML parser and 
> (b) it can't correct any limitations or errors in its parsing.

If MathPlayer is just given a DOM, then all the parsing is done by IE; if 
all the parsing is done by IE, then, again, it's just tag soup, not XML of 
any kind. The term "XML data islands" would be very misleading to anyone 
who supported XML primarily for its strict syntax error handling rules.

So it comes down to a decision between tag soup with namespace prefixes 
and tag soup without namespaces. As noted in earlier mails, the latter 
seems more likely to enjoy a long term success.

Ian Hickson               U+1047E                )\._.,--....,'``.    fL       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Tuesday, 3 October 2006 23:34:42 UTC