[om] Trip Report to MathML Conference

Michael Kohlhase Michael_Kohlhase at sausage.fox.cs.cmu.edu
Thu Oct 26 02:21:34 CEST 2000

Dear Friends,

I have written a trip report to the OMDoc mailing list [1] about my
impression of the MathML conference last weekend at Urbana Champaign (minor
corrections by David Carlisle).

Since this is written from an OMDoc [2] point of view, it might be
interesting to you.  I have appended it below.

The general feeling I took home from the conference is that there is a
large user community, and they are MOVING! Morever, that probably the best
way to input OpenMath is through MathML (I might be wrong here). 

[1] omdoc at mathweb.org; archived at http://www.mathweb.org/~mailists/omdoc
[2] OMDoc: Open Mathematical Documents, uses OpenMath as the formula
    representation language, see http:/www.mathweb.org/omdoc for details

   Dr. Michael Kohlhase,                 School of Computer Science (LTI)
   FB Informatik,  Bau 36, Zi. 220.1     Carnegie Mellon University
   Universit"at des Saarlandes,          5000 Forbes Avenue
   66041 Saarbr"ucken, Germany           Pittsburgh, Pa  15213-3891
   tel/fax: (49)-681-302-4628/5076       tel/fax (1) 412-268-5749/6298
   net: <kohlhase at cs.uni-sb.de>	         net: <kohlhase at cs.cmu.edu>
   http://www.ags.uni-sb.de/~kohlhase    http://www.cs.cmu.edu/~kohlhase

> Dear all, 
> since I was the only one on this list who went, I will write a short (lets
> see) and personally biased summary (for the facts go to the web site
> http://www.mathmlconference.org).
> First of all, the Conference was hosted by Wolfram Research and was very
> well organized. There were about 190 people from, which I found surprising,
> given the frustration over the often-delayed MathML 2.0 standard and the
> lack of visible tool support. 
> I missed the MathML tutorial, so I cannot say much about that. 
> The keynote speaker was Bob Sutor from IBM, who emphasized that the web is
> exploding and that it needs standards and that IBM supports standards (even
> if it has to cooperate with Microsoft for that). Interesting for me was the
> information is that IBM and M$ push the soap spec for secure messaging and
> the UDDI initiative for Internet brokering (see www.uddi.org).
> Then there was a session on the browsers, Internet explorer and
> Mozilla. 
> Dave Massey (Industry type, slick speaker, called M$ file formats like Word
> "not proprietary") presented IE5.5 and gave an online demo. IE integrates
> MathML as external COM objects, which need a special declaration in
> html/head (mozialla ignores that, so it is safe to generate). Even though
> he denied it, MathML is second-class citizen, integration is not good,
> e.g. no searching. MathML COM objects are supplied by two separate vendors
> (Design Science and IBM).
> Roger Sidje gave the talk on Mozilla (totally different type: Hacker,
> inexperienced speaker, problems with setup). But the demo was great. MathML
> is fully integrated with Mozilla, MathML can freely be mixed with HTML,
> images, SVG, you name it. I liked one demo, where he had a moving image as
> the denominator of a fraction (1/<a dragon>), Impressive; made IE look
> pale.
> Then there was Irene Schena from the HELM project of Andrea Asperti at
> Bologna. She was the only one that concentrated on Content MathML with
> DefinitionUrls, uses it for communication with the Coq theorem
> prover. Academically, this is certainly the most interesting project
> presented, (we are talking about cooperation).
> Neil Soiffer from Wolfram presented MathML support in Mathematica. He uses
> presentation MathML to communicate with Mathematica. That is kind of easy,
> since Mathematica also has tree-like formula represenations. Surprising
> thesis is that presentation MathML is sufficient, if it comes from machine,
> i.e. it contains enough <mrow> &invisibletimes; and &ApplyFunction;. Quite
> impressive, but of course only tested on the Mathematica fragment of Math
> (which is of course quite computational and (compared with logic)
> simple). For MathML that is typed in by the user, they use a heuristic
> algorithm to guess the correct formula structure.
> John Plaice spoke about generating MathML from OMEGA (reimplementation of
> TeX). That looks like a good production tool for converting LaTeX documents
> to MathML (he generates PMathML), and (maybe OpenMath, if we can get the
> semantics right). He later told me that it is no problem to get OMEGA to
> generate OpenMath and OMDoc, if the input is sufficiently marked up. I
> think that this is going to be much better than my latex2omdoc.sty. He has
> promised a release on Sourceforge soon (watch
> http://sourceforge.net/projects/omega-system). 
> Then David Carlisle talked about xmltex an XML-parser written in TeX, as a
> consequence LaTeX can directly read in XML (and of course OMDoc).  This was
> not much new to me, since I had played with it (see omdoc/tex/macros/xmltex
> on CVS). As an editor of the MathML specification, he did not have much
> time. He emphasized that xmltex (in contrast to XML parsers) does not
> construct the tree for the whole document in memory, but just traverses the
> input. Therefore it can be much more efficient on large documents.
> Eitan Guari talked about TeX4ht, a program that transforms LaTeX documents
> to XML (html, MathML,...) without changing the TeX kernel. As a
> consequence, (or because it is more complete) many more config files are
> needed. This seemed doable, but  much less elegant than OMEGA. He talked
> about the problems of inferring content from LaTeX, paradigmatic
> examples: for $(a+b)^2$, the TeX parser thinks that the (only the bracket
> is squared what does it care for the content). To get the right kind of
> semantics, we would have to use ${(a+b)}^2$, which of course nobody
> does. OMEGA has the same problems of course  (Junk in -> Junk out).
> I did not see the talk on converting Handwriting to MathML, but that may
> become important for us later.
> The talk about editing MathML with Amaya (the experimental W3C browser;
> open source, see http://www.w3c.org/amaya) was absolutely
> fascinating. Editing in Amaya is seamlessly integrated with browsing (you
> can just place the cursor anywhere (even in MathML of course) and directly
> and intuitively edit the stuff). It would be very convenient if we could
> generated OMDoc that way. I later spoke to Irene Vatton and Vincent Quint
> (the principal developers who gave the talk), they told me that they are
> generating Content MathML with enough &InvisibleTimes; and <mrow> elements
> to make the structure very explicit, they have promised to also put
> &ApplyFunction; generation into the next release (in two weeks). That is
> then enough to recover Content MathML and OpenMath from the output.
> <mrow> bla bla bla &Applyfunction foo bar baz</mrow> goes to 
> <OMA>blublublu foo bar baz</OMA>, where blublublu is recursively 
> generated in the same form (it is the first argument, e.g. another <OMA>).
> This can be done by a style sheet. 
> They told me that the generator is declaratively defined, so that it should
> be possible to directly generate OpenMath from Amaya (I will try once the
> release is out). The only thing that is really missing is the CD names
> then. They could be inserted from a catalogue (kept separately using a
> style sheet), as a kludge. But Irene and Vincent say that it would probably
> be quite simple to extend Amaya (using their Namespace code) with a
> facility that adds DefinitionUrls to <csymbols> (the CMathML version of
> OMS). Then the output of OpenMath would be really simple.
> Finally, they said that it would be relatively simple (but work) to extend
> Amaya with native editing features for docuement-level OMDoc. Then one
> could generate 'xref' and 'for' elements by pointing and clicking. They
> would be prepared to host a C programmer that was interested (that is in
> Grenoble, very nice, anyone interested?) in doing the extension. I see this
> as a very promising path for OMDoc. 
> Robert Miner of Design Science presented an MathML input applet (a
> miniature equation editor) that can be integrated into web pages, and that
> produces Content Mathml. The applet is commercial, but for academic
> purposes, he is prepared to give out source licenses. Robert was interested
> in extending it to maintain DefinitionUrls, and try to generated OpenMath
> output. Might also be interesting to watch. 
> A very differenct but no less interesting talk was by Angel Diaz (old
> OpenMath stock, responsible for TexExplorer at IBM), he showed a system
> that could actually read MathML aloud from web pages, using the structure
> of MathML to generate better output. 
> The most surprising talk was by Paul Karleen of Reed Elsevier. This is the
> publishing company that retypes all the US Patent applications (3000-5000
> per week, imagine that). They have switched to MathML as a representation
> format for the Math in the patent applications. They use a modified version
> of Mathematica to input that. They probably produce more MathML in a month
> than the rest of the world together in a year. 
> Stephane Dalmas talk about using Math Databases for searching Math (using
> automated deduction techniques) was cancelled, since he did not appear
> (neither did his co-author). That was a real pity, since I was very
> interested.
> Stephen Wolfram (second keynote speaker), Jeremy Gray and Patrick Ion (AMS
> math Reviews) gave very scholarly talks about the history of mathematical
> notation and symbols. That was very interesting, but gave me no new
> information for OMDoc.
> There was also a panel session about using Math in web pages. Look at the
> urls to get an impression about them yourself.
> http://www.studyworksonline.com
> http://www.math.niu.edu/~rusin/known-math
> http://mathforum.com -> Dr. Math
> There were a couple of other talks about the use of MathML which I found a
> bit uninspiring (maybe I was just too exhausted), so I will not say
> anything about them. They have promised all the slides to appear at
> http://www.mathmlconference.org in the next couple of weeks).
>      Michael
om at openmath.org  -  general discussion on OpenMath
Post public announcements to om-announce at openmath.org
Automatic list maintenance software at majordomo at openmath.org
Mail om-owner at openmath.org for assistance with any problems

More information about the Om mailing list