Re: correct DOCTYPES to use on MathML 3.0 documents

Frédéric WANG <fred.wang@free.fr> writes:

>     The W3C validator is currently using XHTML1.1+MathML2.0 and
>     XHTML1.1+MathML2.0+SVG1.1, while HTML5 is still indicated as an
>     experimental feature. So I was thinking that a smooth upgrade
>     could be to provide DTDs for XHTML1.1+MathML3.0 and
>     XHTML1.1+MathML3.0+SVG1.1 (or maybe only the latter DTD, which
>     includes the former). I guess DTDs for HTML5 can be generated
>     from the RelaxNG schemas (and it's probably already the case).

There are some things in the text/html serialization of html5 that
would be very difficult, if not impossible, to model faithfully in
SGML.

Html5 rules allowing unquoted attribute value strings are
more liberal than what can normally be enabled with limited shorttag
provisions in an sgml declaration.  For example, with <a href=foo/>
the '/' is understood as part of the value string though with
<mspace width="1em"/> the '/' is understood to make a self-closing
tag because "mspace" is not in the html namespace.

In fact, <a> is defined as having "transparent content", which means
that the content can be any subgrove of html5 content provided it does
not contain an <a> inside.

And as things are now, <a id="thisLocation" /> in text/html is parsed
as <a id="thisLocation"> -- see html5lib at Google code, the python
version -- so that the "adoption agency algorithm" should be invoked
by a user agent to provide a closetag for it somewhere down the road.

The specification does spell out, I think correctly, when the two
character sequence "/>" should be regarded as self-closing syntax, but
then rather strangely flags it as an error and more or less specifies
handling the tag as an opentag when the tagname is in the html
namespace but as a self-closing tag when it is in a foreign namespace.

                                    -- Bill

Received on Tuesday, 12 October 2010 15:34:31 UTC