[whatwg] Mathematics on HTML5

[Henri Sivonen]

On Jun 8, 2006, at 01:18, Ian Hickson wrote:

> On Thu, 8 Jun 2006, Henri Sivonen wrote:
>>>
>>> If we made MathML work in HTML, possibly with rules that make the
>>> syntax easier (by implying tags as I suggested earlier)
>>
>> The implied stuff seems scary. I was hoping for no more tag inference
>> beyond HTML 4 legacy.
>
> Oh? Why?

Because it requires more special work and, therefore, requires a buy- 
in from parser writers and opens up more opportunities for bugs. If  
MathML is generated anyway (by TeX4ht, itex2mml, Mathematica,  
MathType, OpenOffice.org, etc.) and the DOM is complicated anyway, it  
doesn't make much sense to introduce new special cases in the  
serialization.

Also, it would make sense to research if it is feasible to be markup- 
compatible (even if not DOM-compatible) with IE+MathPlayer on the  
text/html side.

> So making MathML+XHTML work in text/html would be a good thing, then?
> (That's basically what I'm proposing here.)

In 2001, when this was on the table, you and I both wanted to go with  
XML media types instead of text/html. Perhaps it is, indeed, time to  
revise the stance here if Roger Sidje suggested it. (On the other  
hand, holding math hostage has slowly shown signs of causing IE 
+MathPlayer change their behavior but the IE release cycle is really  
slow.)

If the WHAT WG defines a way for serializing MathML as text/html, I  
think it should be (at least for conforming cases) a pure alternative  
infoset serialization for a subset of possible infosets (dirty words,  
I know) as opposed to just being something similar but subtly  
different. That is, I think conforming documents should be losslessly  
reserializable as XML *1.0* and the DOM nodes for the math stuff  
should report the uncolonified element name in lower case as  
localName and the MathML namespace URI as namespaceURI. (I guess  
tagName should do what is most compatible with MathPlayer.)

Finally, Mozilla's HTML parser and content sink are a mess. For  
sanity and interop, it would be good to get those reimplemented per  
spec rather than having new MathML stuff thrown into the current mess.

-- 
Henri Sivonen
hsivonen at iki.fi
http://hsivonen.iki.fi/