I have two opinions regarding xHTML. First, xHTML should not be used for public consumption. Second, xHTML should be used for public consumption. The latter stems from the original vision of xHTML and
the W3 recommendation that the World Wide Web transition away from HTML to XML/xHTML. The former stems from this failure.
HTML 3 was a failure, too. It was realized that markup for visual
presentation made information inaccessible to non-visual parsers
(and by extension, the blind,
and in some cases, the hearing impaired). The strength of HTML 3 was
backward compatibility to IETF HTML. HTML 4, therefore, in a large effort to also be
backward compatible, offered the Transitional document type
definition that could be used for transitioning HTML 3 documents to 4.
The expectation was that new documents would be written in Strict.
HTML 4 also introduced frames, a method of embedding multiple HTML
documents into a single view. This latter technology, as with the
former visual markup inherent to HTML 3.2 in Transitional, was
rejected when finally formalized as an ISO standard (15445:2000E).
The same form of thinking was attempted with xHTML. Though certain
elements of HTML 4 were deprecated in xHTML 1.0, and certain
deprecated elements of HTML 3.2 removed, a backward compatibility was
attempted with xHTML 1.0 that would transition HTML 4 documents to xHTML. As with HTML 3.2, this was a failure. The difference is that what xHTML offered (XML) was not
recognized, and it did not present any new features (other than what was inherent to XML) from HTML 4.
The problem with xHTML 1.0 is not inherent, but relative to the need
for xHTML/XML user agents to parse xHTML pages. The backwards
compatibility was lost.
What made this worse is the informative suggestion of xHTML 1.0's
Appendix C. This suggested that xHTML could be served to HTML user
agents, even though xHTML was incompatible with HTML, because
existing web browsers (which were and are the most popular form of
user agent) already had built in error correction facilities that
would overlook the incompatibilities as tag soup syntax errors. If
the MIME types note is adhered to, the implication of
Appendix C becomes a reliance on the HTTP protocol capability to
send user agents what document types were
available (and user agents to advertise what they could use), so that server negotiation could be used (such as Apache's
MultiViews directive) to send an xHTML document as text/html for HTML user agents, and as
application/xhtml+xml
for xHTML and XML user agents, respectively. The problem is that
though the W3 front page does this, nearly no one else does, even some
W3 projects.
The W3, and Tim Berners-Lee, have decided to slow down the approach to
an XML only world wide web. The future is to deprecate xHTML 1.0 with new
recommendations that will attempt to replace xHTML 1.0 with xHTML 5,
while replacing HTML 4.01 with HTML 5. HTML 5 will have the graceful
upwards compatibility to xHTML 5 that was originally hoped for with
xHTML 1.0 and backwards compatibility with HTML 4.01. This will
allow an intermediate process of adoption of XML technologies, while
still allowing user agent developers, such as those for the HTML only
Internet Explorer web browser, to continue as they are. The xHTML2
Working Group will continue to build on the work found in xHTML 1.0
and 1.1.
This is a more realistic model. Standards recommendations should
reflect what people do well, not what standards bodies want people to
do. The problem with xHTML 1.0 was a lack of foreseeing that not
every one would adopt XML immediately, or understand how it is
different. Appendix C of the xHTML 1.0 standard informs
developers, and web server administrators, to send xHTML 1.0 documents
to non-xHTML user agents, but with the assumption that the world would transition to xHTML. Except that Microsoft's Internet Explorer, as of IE8 beta 2, is
still working on support of HTML technologies, and most developers
adopted xHTML as if it were a HTML technology, sending xHTML as
text/html to everyone, and neglecting entirely XML user agents. This
has the very opposite effect that xHTML 1.0 was intended to achieve.
As much as my preference is for the adoption of xHTML, for now, the
best solution seems to be to abandon xHTML 1.0 for public consumption
(only sending it with the correct MIME type when used, and abandoning the
Appendix C model), use xHTML technologies (Ruby, XForms, MathML, etc.) where its XML capabilities
are necessary (such as in embedded devices), and use the existing HTML
4 recommendations, and the ISO standard based on it, for the Desktop using
public until such time as HTML only user agents are no longer prevalent.
