102244 - MathML shouldn't require filenames ending in .xml

Status: VERIFIED INVALID

(Core :: MathML, defect)

Description

[Ken Harris]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.4) Gecko/20010918
BuildID:    2001091809

MathML in Mozilla seems to only kick in for (1) files ending
in .xml or .xhtml, (2) start with <?xml version="1.0"?> (i.e.,
are really XML), AND (3) have a <!DOCTYPE ... "mathml.dtd">.

It should kick in whenever it sees <math>, even for files that
end in ".html".  Otherwise authors who want to use Mozilla's
MathML support have to call their files this.xml or that.xhtml,
which don't seem to render at all in other browsers I've
tried (Amaya, IE5).


Reproducible: Always
Steps to Reproduce:
1. Download and install a -MathML build (like 0.9.4-MathML)
2. Go to http://www.w3.org/Amaya/User/Math.html


Actual Results:  Math markup isn't rendered as such.

Expected Results:  It should be rendered as math.

The only other MathML-enabled browser I could find is Amaya.
The URL listed above is a file called "Math.html", but it's
XHTML: a decent enough compromise, I think, since
non-MathML/XHTML browsers can still display most of it as
intended.  (Amaya renders it correctly, including the math.)

Comment 1

[Christopher Aillon (sabbatical, not receiving bugmail)]

Confirming on linux CVS pull from the 14th.

->XML.

Comment 2

[Heikki Toivonen (remove -bugzilla when emailing directly)]

I think this is either INVALID or WONTFIX.

MathML is an XML language, therefore we should do nothing special about MathML
markup if we are served HTML or any other non-XML mime type. Please not that the
file suffix is not the issue here, it is the mime type. A server probably maps
.html to text/html mime type, and .xml and .xhtml to text/xml or some other
standardized XML mime type.

I would be somewhat surpised if the XML declaration and the MathML DTD were
required, though.

Comment 3

[rbs]

The MathML DTD is needed to provide access to the numerous MathML entities
(there are over 2000 of them - no way they can be in the internal set). That's
why I consider bug 44458 of great importance because its framework might provide
a hook to support the official MathML DTD in a portable manner (instead of the
current non portable hack: <!DOCTYPE ... "mathml.dtd"> which breaks other
browsers because it isn't in their local/system list of DTDs).

In the meantime, marking this bug INVALID (the mime-type is ultimately the
deciding factor - see also bug 67646 about the issues involved in relaxing
this).

Comment 4

[Hixie (not reading bugmail)]

VERIFIED -- we do trigger MathML for *.html documents, so long as they are sent
as text/xml. MathML is, as mentioned above, an XML vocabulary and therefore
should only be recognised in an XML context.

Comment 5

[Ken Harris]

The "INVALID or WONTFIX" comment scares me.  From the "bug's life cycle"
document, this means, basically, "We don't know if this is a bug or not,
but we're not going to do anything about it."  (Why is a decision being
made not to fix a bug before it's decided if it even is a bug?)

Why this should be fixed:

- That's how lots of existing web pages are.  w3.org's own MathML samples
are sent out as MIME type text/html.  (Can you point me to a site that
actually sends out MathML in an XHTML document as text/xml?)

- Other MathML browsers don't have this restriction.  Amaya handles
MathML in text/html files just fine.

- Users want this fixed: see
http://www.openwiki.com/ow.asp?a=view&p=HelpOnMathML&redirect=MathML

- Maybe I'm completely missing it, but I don't see anywhere in the MathML
spec that says anything about MIME types, especially something as drastic
as "if it's not served as text/xml, you may completely ignore MathML".

- Users don't necessarily have control over this.  In the simplest case,
if I'm not the web server administrator, I can't control text/html versus
text/xml.  But also any program that generates HTML files (like Javadoc)
is automatically excluded.

So I can think of 5 reasons this should stay open.  The only reason to
close it I've seen is "MathML is an XML vocabulary".  That's a good
theoretical reason at best (though I can't find where in the spec it says
this).  Is that really more important than being able to render more web
pages as they were intended?

Comment 6

[Heikki Toivonen (remove -bugzilla when emailing directly)]

The resolution is INVALID, which in our opinion means "The problem described is
not a bug". If you would file an enhancement request "Want MathML support in
HTML" it would be marked as WONTFIX (or maybe Future, which might actually end
up meaning the same thing).

W3C's webserver is sending a lot of rubbish anyway; you can't trust them to do
the right thing with their server (for example, W3C servers can serve you the
raw XML document of a specification without any stylesheet if your browser
accepts any XML mime types). mozilla.org hosts some MathML examples with an XML
mime type.

Users typically have the power over the file names they upload to a webserver.
It  should be quite easy to get the server administrators to support some XML
file suffix to XML mime type mappings if they don't do that already. If you can
run CGIs you have complete control over the mime types anyway.

"MathML is an XML application" from the Abstract of MathML 2.0 recommendation at
http://www.w3.org/TR/MathML2/

The conformance section of the said specification points to a W3C test suite
which contains only pure MathML XML test documents and XHTML test documents.

Appendix A of the said specification discusses Parsing MathML. It refers to
MathML XML DTD for validation, talks about namespaces and specifically says in
"A.2 MathML as a DTD Module": "[MathML] is designed to be used as a module in
documents marked up with the XHTML family of markup languages".