W3C home > Mailing lists > Public > www-html@w3.org > March 2002

Re: filename ending - .html or .xhtml

From: William F. Hammond <hammond@csc.albany.edu>
Date: 13 Mar 2002 18:50:41 -0500
To: mozilla-mathml@mozilla.org, www-html@w3.org
Message-ID: <i71yeohvse.fsf@pluto.math.albany.edu>
Chris Hoess <choess@force.stwing.upenn.edu> writes to
mozilla-mathml@mozilla.org:

> In article <3C8FB170.3010304@spth.de>, Philipp Krause wrote:
> > I wrote a XHTML page that uses MathML, saved it with the ending .html;
> > Mozilla interpreted all the XHTML stuff correctly, but not the MathML.
> > When I changed the filename to one with the ending .xhtml it worked ok.
> > Shouldn't Mozilla ignore the filename and simply interprete the content?

As things are "application/xhtml+xml" really ought to get the idea
across.  It _appears_ to have the added advantage of sponsorship by
key players in the HTML WG.  However, not all user agents of general
market importance will handle MathML in XHTML this way.

While Mozilla appears to handle MathML in XHTML under the types
"text/xml" (which in some minds should not exist) or "application/xml",
it is ridiculous to perceive Mozilla or any other broad market user
agent as fully capable of handling the scope of these types (see
RFC 3023).  This gives rise to concern that such use of these types
for html family documents will have the effect of making these types
useless for XML instances not in the html family.

In fact, there is a dilemma for a rigidly correct text/html user
agent.  RFC 2854 leaves open in some minds the possibility, without on
the other hand providing authorization, that text/html is an
acceptable mime type for this.  Even if that is not acknowledged there
is scant justification for an XML-capable rigidly correct user agent
rolling over something it's obviously capable of handling.  The reason
Mozilla does this is that its design provides efficient handling by
making processing decisions based solely on mime level information
before the body of the http object is seen.  So to be rigidly correct,
it would need to restart itself.

One way around this would be to ask for allowing the header

        Content-type: text/html; profile=xhtml

(i.e., allow "profile" as a content-type parameter for "text/html") in
a suitable public spec.  A content provider using something like this
would need to be aware that an "old" user agent would ignore the
parameter.  There are various ways that such dual usage for a single
http object could be facilitated.

                                    -- Bill
Received on Wednesday, 13 March 2002 18:51:32 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 March 2012 18:15:50 GMT