109825 - Allow scripts to detect MathML-enabled builds

Status: RESOLVED WONTFIX

(Core :: MathML, defect)

Description

[rbs]

Filing this bug for the record. There was thread on n.p.m.mathml asking how 
scripts can detect MathML-enabled builds. The thread is: 
http://groups.google.com/groups?hl=en&th=6ceed7fc31a0262e&rnum=1

When MathML becomes an integral part of Mozilla, it is conceivable that this
might still be useful since some embedders may choose to disable it.

Following the code done in nsGenericElement::InternalIsSupported(), the 
fix for this could simply be to add recognition of something along the lines:
DOMImplementation.hasFeatures(feature = "MathML", version = "" | 1.0 | 2.0).

This is in line with the spec:
http://www.w3.org/TR/2001/WD-DOM-Level-3-Core-20010605/core.html#ID-102161490

Comment 1

[Johnny Stenback (:jst)]

nsGenericElement::InternalIsSupported() doesn't seem like the right place to put
this, that method should be for *DOM* features only, there's no MathML DOM is
there? I'm happy to expose this *somewhere* in the DOM, but I'm not sure the DOM
feature query mechanism is the right place for it.

Comment 2

[rbs]

There is now a MathML DOM (in the latest V. 2.0):
http://www.w3.org/TR/MathML2/appendixd.html

There is a gray area in distinguishing between the MathML-DOM (lately introduced
in v2.0) and MathML-layout (which is all that was in MathML 1.0 that we aim at
for now). I have been wondering myself about the suitability of
DOMImplementation.hasFeatures() for both cases. The DOM as known today wasn't
yet there when MathML 1.0 was sitting there as an unimplemented recommendation
in mainstream browsers.

Comment 3

[Andreas Franke (gone)]

I think this would be nice to have for 1.0, because a significant number of
builds from that branch will have mathml enabled.

Comment 4

[rbs]

mozilla.org itself is shipping builds that may or may not be MathML-enabled (Mac
vs Others).

jst, did you got a definite idea where the detection should go?

Comment 5

[Hixie (not reading bugmail)]

rbs: Do we support the MathML DOM?

Comment 6

[rbs]

No, that's why it is uncertain how to fix this bug.
(Note that the proposed MathML DOM cited in comment #2 is non-normative) 

Comment 7

[rbs]

I had this idea of leveraging on document.load()... but curiously, the security
manager won't let me use the data: protocol. Will attatch the test I did.

Comment 8

[rbs]

Attachment: testcase to check if MathML is enabled

The idea is to create a XML document and load some of those weird MathML
entities.
And if there is no parsing error during load, conclude that MathML is there.

Comment 9

[rbs]

I am wondering if there are security risks in doing:
  document.load("data:text/xml, some-weird-entities-from-the-MathML-DTD")
if the security manager allowed the data: protocol, it have been a good enough
fix for the time being.

Comment 10

[rbs]

Filed bug 201903 about the fact that document.load("data: protocol) doesn't
work. With that, this bug could have an interesting standard-compliant solution.

Comment 11

[Henri Sivonen (:hsivonen)]

All MoCo builds now have MathML enabled, right? Would it be appropriate to WONTFIX this now?

Comment 12

[Frédéric Wang (:fredw)]

(In reply to comment #11)
> All MoCo builds now have MathML enabled, right? Would it be appropriate to
> WONTFIX this now?

Yes, I agree. Moreover if we consider other browsers, Opera supports MathML and implementation is underway for WebKit (Internet Explorer still requires a plugin though). I suggest closing this bug if nobody complains.