Re: //head/object

"Toby Inkster" <tai@g5n.co.uk> wrote:

> On Tue, 4 May 2010 16:20:22 +0200
> Krzysztof Maczyński <1981km@gmail.com> wrote:
> 
> > I can't see how it makes those browsers' parsers simpler.
> 
> Certainly allowing <object> in <head> doesn't increase parser
> complexity. It would be a simple matter of switching from treating it
> like, say, <div> which automatically triggers the <body> element to
> start, to treating it like <script> which does not.

It's not quite that simple. <script> doesn't allow child elements, so it doesn't illustrate the issue. Instead, consider the only possible child of <head> that can have child elements of it own: <noscript>. It has a tree builder mode of its own to deal with the special cases that arise.

It seem to me that to work IE-like <object> into the HTML5 tree builder, <object> would have to switch to "in body" and </object> would have to reset the insertion mode to "in head". I haven't worked out what unexpected consequences this would have.

"Neil Soiffer" <NeilS@dessci.com> wrote:
> IE's binary behavior's make use of <object> inside of <head>:
> http://msdn.microsoft.com/en-us/library/aa744098%28VS.85%29.aspx
> 
> Breaking this compatibility with HTML4 will break all of the existing
> web
> pages that make use of this facility, 

Binary behaviors are an IE-only feature, so I'd expect such pages not to be compatible with Web browsers in general to begin with. Do you have examples of Web pages that use binary behaviors and work reasonably in browsers other than IE?

While the HTML5 parsing algorithm aligns with some IEisms for the purpose of recovering from minor errors on Web pages that try to use cross-browser markup, supporting deeply IE-only legacy features, such as conditional comments, custom tags or XML data islands, hasn't been a goal for HTML5.

> including the users of our
> MathPlayer
> MathML renderer for IE.  I did an experiment and put the object tag
> inside
> of <body> and MathPlayer was still triggered, but existing pages would
> not
> work in HTML5.  Our usage stats say tell us that over 1.5 million
> pages with
> MathML are being viewed with MathPlayer per month, so breaking that
> compatibility will affect plenty of people/content.

How do those pages work for Firefox 3.6 and Opera 10.5x users? If the pages are served as text/html to all browsers, the pages are already broken in shipping non-IE browsers (and even in IE without binary plug-ins). If they are served as application/xhtml+xml, the HTML5 parsing algorithm doesn't apply to them, so things are OK. Am I missing some consideration here?

Do you have examples of typical pages that use MathPlayer?

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

Received on Wednesday, 12 May 2010 10:56:55 UTC