W3C home > Mailing lists > Public > www-math@w3.org > October 2006

RE: MathML-in-HTML5

From: Ian Hickson <ian@hixie.ch>
Date: Wed, 4 Oct 2006 19:57:12 +0000 (UTC)
To: Robert Miner <robertm@dessci.com>
Cc: "Roger B. Sidje" <rbs@maths.uq.edu.au>, www-math@w3.org, dev-tech-mathml@lists.mozilla.org, David Carlisle <davidc@nag.co.uk>
Message-ID: <Pine.LNX.4.62.0610041942500.25753@dhalsim.dreamhost.com>

On Wed, 4 Oct 2006, Robert Miner wrote:
> 
> The main benefit to allowing namespace prefixes on the math elements in 
> HTML5 is interoperability with Internet Explorer.  This is a key point, 
> so I hope you will forgive me if I explain at length.

But we won't get that. IE's processing model is massively involved and far 
more complicated than I'd even remotely like to specify (I've been 
involved in at least two browser vendors over the years who have run afoul 
of either trying to emulate IE's behaviour here or trying to avoid having 
compatibility issues because of it).

Furthermore, over the 6+ years that IE has supported this model, it has 
gained almost zero traction. Why would we copy a model that has so clearly 
demonstrated failure?


The argument that we must use the IE namespace model here seems to rest on 
the idea that this is required for compatibility with MathPlayer.

Why couldn't MathPlayer change just like Mozilla is changing?

Web Forms 2.0 (the forms extensions being added to HTML5) are being 
implemented in multiple browsers natively, and are being implemented in IE 
using JavaScript shims.

Canvas (another HTML5 feature) is implemented in all the major browsers 
except IE natively, and enjoys IE support through, again, a JavaScript 
shim.

Why wouldn't MathML support work the same way? It could be a JS shim that 
interacts with MathPlayer, or just one that works on its own. It could 
also just be MathPlayer on its own hooking in to every page -- it doesn't 
_need_ to be done via the proprietary namespace dispatch mechanism.


As far as I'm concerned, anything that requires a namespace prefix 
mechanism, or any namespace declaration of any kind, is fundamentally 
flawed and doomed to have near-zero market uptake. There is ample evidence 
of this on the Web. All the support forums for technologies like XHTML and 
browsers that support namespaces are full of people asking questions about 
namespaces and being generally confused by them, and the IE mechanisms 
have themselves got no serious traction on the Web outside of very 
specific fields (like Office's "HTML" document round-tripping feature).

Furthermore, any mechanism that is based on IE's namespace extension 
mechanism will cause *huge* headaches for browser vendors, because there 
is a massive amount of content on the Web that incorrectly uses these 
mechanisms in ways that are very specific to IE's error handling and 
namespace registration model. There are literally tens of millions of 
pages, for instance, that use the "xmlns" attribute on the <a> element, 
and even more that do so on the <br> element. Short of making the new 
parsing rules ridiculously complicated, it would be impossible to get 
decent compatibility with existing content.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'
Received on Wednesday, 4 October 2006 19:57:25 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 20 February 2010 06:12:59 GMT