PDF version (509 KB)
Math on the Web: A Status Report
Focus: Adding Value for STM Publishing
by Robert Miner and Paul Topping, Design Science, Inc.
Over the last half year a tidal shift has taken place in the state of math on
the web. High-quality math support is now available in Netscape, Mozilla and
Internet Explorer browsers. In August, Netscape 7  was released
with native math support on the Unix and Windows platforms. Math support on the
Mac platform came about the same time in Mozilla 1.1 , the open source sister
browser of Netscape 7. MathPlayer , a free
extension for Internet
Explorer  from Design Science, completed the field in September by adding
native-quality math support to the web's most widely deployed browser.
Better browser support for math has significant implications for Scientific,
Technical and Medical (STM) publishers. Most scientific web publication is
presently done using PDF
. Now it is also feasible to effectively publish technical content in
HTML + MathML format, which may turn out to be cheaper and better in many
situations. In response to increased interest, several new MathML-capable tools
have been announced catering to the STM market. This edition of the Status
Report takes a closer look at these tools.
The HTML + MathML Platform
The technology underlying the new browser support for math is MathML , a World Wide Web Consortium (W3C)  Recommendation
for encoding mathematics in XML format. MathML is designed to be used in
conjunction with some other document-level markup language, since it is only for
encoding mathematical notation. For web publication, the natural document markup
language to use is XHTML, the XML-compatible version of HTML. In the publishing
arena, DocBook and a few other XML-based document markup languages are also
Of course, a modern web page consists of much more than static HTML code. Web
pages are frequently dynamic, with features like menu highlighting implemented
sheets to keep content and visual presentation separate for easier maintenance
In past editions of the Status Report, we have dubbed HTML, MathML and
related web technologies the HTML + MathML platform. Conceptually, the HTML +
MathML platform is made up of three parts -- markup languages for encoding
content, stylesheets for controlling how content is displayed, and a programming
model for making the page dynamic. The primary markup languages involved are HTML , XHTML , MathML, RDF  for metadata and SVG  for structured graphics. The
most important stylesheet languages are XSL  for transforming XML into HTML,
and CSS  for visual style
information. For web programming, the Document
but many other technologies may be involved, such as applets, plug-ins and so
Standards work at W3C and browser implementation have been converging toward
the HTML + MathML platform for some time. With MathML support in major
browsers under Windows, Unix and MacOS, the HTML + MathML platform achieves a
new level of viability. Moreover, because all the component technologies were
designed around a common XML-based conceptual framework, they work well together
in XML-based workflows. XHTML + MathML web pages now offer a strong alternative
to PDF for many kinds of technical web publication. As a result, there has been
a surge in MathML interest and activity as people take a closer look at the new
MathML Gains Momentum
MathML is relatively venerable as web standards go. It was released in 1998,
just a couple of months after XML itself was completed, and before CSS2 was
available. In the intervening time, a number of scientific software packages
have added support for MathML, and several applets and plug-ins have long been
available for displaying MathML in browsers. However, these early display
technologies were not robust enough or well enough integrated with browsers to
be a viable option for demanding applications. So even though interest in better
solutions for math on the web remained high, most individuals and organizations
adopted a wait-and-see attitude after initially investigating MathML.
But now the wait is over, and people are liking what they see. Unlike earlier
technologies, Netscape/Mozilla and MathPlayer are getting high marks in the
marketplace. MathPlayer was chosen as a Hot
Pick  at the Seybold 2002 conference for professional publishing.
Judging by their initial reception, MathPlayer and Netscape/Mozilla promise to
change the landscape for math on the web dramatically.
In Netscape 7 and Mozilla 1.1, MathML support is built into the rendering
engine. In speed and quality, it is comparable to the rest of the browser text.
Because it is built in, users don't need to download a separate plug-in.
However, many users find they need to download and install math fonts.
In Internet Explorer, MathPlayer provides MathML support. MathPlayer utilizes
powerful, low-level extension capabilities called behaviors only
available in the Windows version of Internet Explorer. However, by utilizing
behaviors, MathPlayer achieves high-performance, native-quality rendering and
seamless browser integration. MathPlayer is installed by downloading a standard
Windows installer. The installer also includes the fonts needed by
Even with much improved support for MathML in browsers, some technical
challenges still remain. Because of differences in the way in which
Netscape/Mozilla and Internet Explorer handle XML documents, in practice many
people find it necessary to publish HTML + MathML documents with an XSL
stylesheet that customizes the document to the browser. Nonetheless, the new
math support has recently spawned a variety of interesting, experimental
projects. Two representative examples are an online formula finder  and an open source MathML stylesheet archive
. Postings to newsgroups indicate that although dealing with browser
differences is still a painful subject, the technical problems are
Of course the acid test for MathML support in browsers is whether it is
adequate for large-scale publication of technical information, such as
scientific journals. To be credible as a solution in that arena, a candidate
technology has to demonstrate that it looks good, renders fast, and prints well
in a browser, even for long, dense, research articles. But indications are good
that MathML support in browsers now largely achieves that goal.
Conventional wisdom holds that investment in new technology drops off in a
down economy. Consequently, the mere fact that the HTML + MathML platform in now
feasible is not what has attracted attention from STM publishers. Rather, it is
because MathML is both information-rich and XML-friendly, and thus it presents a
number of enticing possibilities for cutting production costs and adding value
to technical documents.
New Interest in MathML from STM Publishers
All for-profit businesses seek to cut costs and increase sales. For STM
publishers, MathML has appeal on both fronts. Publishing is a very mature
industry, and thus finding ways to innovate and differentiate a brand or product
is a major challenge. On the surface, the prevalence of the web offers
publishers a fertile new arena to work in. However, for many STM publishers
dealing with highly technical material, the web has been a problematic medium in
practice. Users easily come to take for granted online versions of articles and
books, and are reluctant to pay extra for them. At the same time, because of
lack of browser support, producing online versions of articles with a lot of
math in them is very expensive, frequently involving a second, independent
workflow for web publishing in parallel to the main print workflow.
In order to address the high cost of web publication, many publishers are
moving toward XML-based workflows, where the same document can be composed as
PDF for print and as HTML for web publication. In this context, the appeal of
MathML for STM publishers is obvious. By using MathML to encode equations, XML
documents can be self-contained. There is no need to generate and store hundreds
of images of equations along with a document. Further, since MathML is an XML
application, documents can be uniformly processed using industry standard tools
such as XSL stylesheets. In the past, it was often necessary to somehow extract
the math for separate processing, and then merge it back into the text later in
the composition process. For a more detailed analysis, see the Design Science
white paper MathML
Workflows in STM Publishing .
The short term cost-cutting benefit of unifying workflows is noteworthy. But
MathML's potential for adding value to web publication may be even more
significant. In addition to fitting nicely into XML workflows, MathML is an
information-rich way to encode mathematics. It takes pains to insure the
hierarchical structure of the markup coincides with the mathematical structure
of the expression. As an example, in the expression (x+2)^2, the MathML markup
structure makes it clear that the exponent applies to the entire expression, not
just the final parenthesis. MathML also provides a means of directly specifying
the mathematical content of an equation in markup in addition to the
presentation markup that describes how an equation should be typeset.
Because there is so much information in a MathML expression, it can be used
in ways that are impossible for the equivalent print expression. For example,
MathML equations can transfer between applications using cut and paste. A
researcher might cut a MathML equation from a web browser, and paste it into a
computer algebra system such as Mathematica
 or Maple . Or a student
could paste an equation into an interactive graphing applet like the WebEQ Graph
Control. Accessibility is another area where information-rich MathML might play
a significant role. MathML was designed with a view to voice rendering for the
vision impaired. Accessibility legislation requires many commercial and
governmental organizations to publish material in accessible format when
possible. There have been a few prototype projects, and there is considerable
synergy with other technologies such as VoiceXML . For a more comprehensive look
at the possibilities MathML offers, see the Design Science white paper
Adds Value to STM Publishing .
Before you can do any of these slick new things with an equation, you have to
find it. Fortunately, because the presentation and meaning of an equation are
tied together by the markup structure in MathML, it has great potential for
improved searching and indexing of technical material. As increasing numbers of
documents containing MathML appear on the web, metadata for math will become
increasingly important as well. The timing is good for increase math metadata
activity, particularly since there are signs that standards and technologies for
handling metadata in general are beginning to stabilize. For example, Adobe has
begun a major initiative to deploy a common way of storing and accessing
metadata. Also, several metadata standards have been successfully employed for
some time in particular vertical markets such as NewsML  for newspapers and PRISM  for magazine articles.
While MathML offers significant potential both for cost reduction and adding
value, one might justifiably counter that it is unwise to count your chickens
before they hatch. While a number of STM publishers are working on MathML-based
projects, there are not yet many large, high-volume, integrated XML workflows
incorporating MathML. To a large extent, this is a matter of inadequate support
in the high-end tools.
Significantly, the tool situation has begun to change in response to customer
demand. Since demand ultimately determines the success or failure of a
technology, new demand for tools is worth a closer look. In the following
section, we will focus on forthcoming MathML support in several important XML
and HTML tools.
Focus: MathML Support in Web Publishing Tools
MathFlow and PTC
The most ambitious integration of MathML support with high-end publishing
 to date is a partnership between Design Science and PTC.
MathFlowTM for Arbortext combines aspects of Design Science's
WebEQ  and MathPlayer
products to provide comprehensive MathML functionality for PTC's
Arbortext Editor  and backend E3 e-content engine. MathFlow for Arbortext was announced at XML World
2002 in December, and is currently in beta testing.
MathFlow utilizes a combined DocBook and MathML markup language called
AxDocBook + MathML, which extends Arbortext's standard DocBook support. MathFlow
consists of three parts. MathFlow Exchange works with Arbortext Editor's Import/Export
to import documents from Microsoft Word containing MathType equations. Equations
are converted to MathML while the surrounding document is converted into AxDocBook.
Once an AxDocBook + MathML document has been opened in the Arbortext Editor, both
the math and the document can be edited naturally. Equations appear in typeset
form. Clicking on an equation opens it in the MathFlow Editor. Closing the
Editor reinserts the typeset equation into the document. In general, editing in
Arbortext Editor is reminiscent of word processors, and the feel of the
integration will be familiar to Word/MathType users.
Once a document is finished, MathFlow and Arbortext Editor work together to generate both
PDF and web output. To compose a document as PDF, users can choose from either XSL or FOSI stylesheets, which are used to transform AxDocBook + MathML into a
low-level composition language used for formatting documents. The math equations
are rendered into PostScript by the MathFlow Composer. The typeset equations and
the remaining formatting code are then combined and converted to PDF by the
For web output, there are two options. Users can use an XSL stylesheet to
convert AxDocBook + MathML into XHTML + MathML for use in new browsers.
Alternatively, Arbortext Editor/MathFlow can generate Design Science image-based MathPage format,
looking web pages that print at 300dpi. MathPage documents extend accessibility
back to the older 4.x browsers.
Dreamweaver and WebEQ Author
While MathFlow and Arbortext Editor are high-end tools aimed primarily at corporate
users, Adobe Dreamweaver
 and WebEQ Author are for a more mainstream audience.
Dreamweaver is a widely-used HTML editor and site development tool. WebEQ Author
adds MathML support to Dreamweaver in a way analogous to that in which MathFlow
works with Arbortext Editor.
Installing WebEQ Author adds an equation editor button to the Dreamweaver
toolbar. Clicking the button opens the WebEQ Editor where an author creates an
equation. Closing the editor inserts a preview of the equation in the
Dreamweaver editing window. Double clicking the preview reopens the equation in
the WebEQ Editor.
Equations are encoded as MathML code in the HTML configured to display
properly in Internet Explorer with MathPlayer. However, because Dreamweaver
doesn't support XHTML, only HTML, web pages created with Dreamweaver can't take
advantage of the MathML support in Netscape/Mozilla without further editing.
Since cross-browser compatibility is often important, WebEQ Author also lets
authors generate web pages where equations use the image-based MathPage format
Although XHTML has an advantage for cross-platform interoperability, HTML has
advantages of its own. Most notably, HTML pages have much better support for
interactivity in browsers. To take advantage of that, WebEQ Author includes a
Solutions Library of templates for interactive mathematical web pages such as
online quizzes, interactive graphing and plotting, and online tutorials. The
templates utilize both dynamic web program techniques, and MathML-aware applets
to provide graphing, evaluation, and equation editing capabilities within in a
MathML-aware applets that go into the templates will also be included in version
3.5 of the WebEQ Developers Suite, which will also be released in early
Filling in Workflow Gaps
The MathFlow/Arbortext Editor and WebEQ Author/Dreamweaver combinations are significant
because taken together with MathType/Word, they provide end-to-end MathML
workflow solutions where none have previously existed. However, they will not
remain alone for long. Other vendors are also moving to fill in remaining gaps
in XML + MathML workflow tools.
Plans have been announced to develop a version of MathFlow for Corel's XMetal  editor. XMetal
has an import from MS Word feature and supports word-processor-like editing of
XML documents. It also has basic printing functionality, though typically in
workflows XMetal is used in conjunction with other composition engines such as
XyEnterprise XPP  which has also
recently added MathML support.
While workflow tools such as MathFlow and Arbortext Editor compose to PDF, many book and
magazine publishers use QuarkXPress for composition. There are several mature
math plug-ins for Quark of which Powermath is perhaps the most well-known. None
of them currently have MathML support. However, a conversion tool, MathMonarch  from
Westwords Publishing, can help bridge this gap.
MathMonarch 5.0, currently in beta testing, can do two-way translation
between MathType Equation Format, MathML, LaTeX and WWDoc, the math markup
language used by Powermath. MathMonarch launches from the toolbar in MS Word,
and displays a control panel where the user specifies the input and output
formats. Equations are converted to the desired format in place in the Word
document. Other tools must be used to process the non-math portions of the Word
document into other formats such as Quark's format or XML.
At present, there are a number of third-party software packages that address
the problem of converting MS Word documents into XML. At least one,
eXtyles from Inera, Inc , uses Design
Science's MathType technology to convert Word equations to MathML.
However, the upcoming release of Microsoft Office 11 will have a large impact in
this area, since support for XML is a major new feature in Office 11.
Consequently, it seems likely that support for MathML in Word to XML conversion
will remain somewhat ad hoc in nature until the dust from the Office 11 release
This section spotlights important developments that have been announced since
the most recent edition of the Status
Report  was published in September 2002. The list may not be complete,
and the authors apologize in advance for any omissions.
- The MathML Handbook is published. Charles River Media has published
a book by Pavi Sandhu on MathML. The book provides a primer of MathML
concepts, discusses techniques for working with MathML, and provides reference
- MathFlow for Arbortext Announced
. Design Science announced its MathFlow for Arbortext product at
- WebEQ Developers Suite 3.5 beta released. Beta testing for WebEQ
Developers Suite version 3.5 began in December 2002. New features include a
MathML-aware graphing applet, MathML evaluation capabilities, and templates
- MathPlayer is
Pick . MathPlayer, Design Science's high-performance MathML
rendering behavior for Internet Explorer was chosen as a Hot Pick at the
Seybold 2002 exhibition in San Francisco.
- XSLT MathML Library Version
2.0 . An open source project to develop XSL stylesheets to convert
from MathML to LaTeX was launched at Source Forge.
- New version of MathML Test Suite released. The official MathML Test Suite has be expanded
- OMDoc mode for Emacs released. A beta version of an extension to
the popular Emacs editor was released as part of the CCAPS project at Carnegie
Mellon University. When completed, the OMDoc "mode" for Emacs will contain
support for editing MathML.
 Netscape 7.0, http://channels.netscape.com/ns/browsers;
Mozilla 1.1, http://www.mozilla.org/;
Microsoft Internet Explorer, http://www.microsoft.com/windows/ie/default.asp;
Adobe's Portable Document Format (PDF), http://www.adobe.com/acrofamily/main.html;
World Wide Web Consortium (W3C), http://www.w3.org/;
Markup Language (HTML), http://www.w3.org/MarkUp/Overview.html;
Extensible Hypertext Markup Language (XHTML), http://www.w3.org/MarkUp/Overview.html;
Resource Definition Framework (RDF), http://www.w3.org/RDF;
 Scalable Vector
Graphics (SVG), http://www.w3.org/Graphics/SVG;
Extensible Stylesheet Language (XSL), http://www.w3.org/Style/XSL;
Cascading Style Sheets (CSS), http://www.w3.org/Style/CSS;
Document Object Model (DOM), http://www.w3.org/DOM;
 Seybold 2002 Hot
 XSLT MathML Library Version 2.0, http://xsltml.sourceforge.net/;
MathML Workflows in STM Publishing,
Wolfram Research (Mathematica), http://www.wolfram.com/;
Maple (Maple), http://www.maplesoft.com/;
 VoiceXML, http://www.voicexml.org/;
Adds Value to STM Publishing, http://www.dessci.com/en/reference/white_papers/mathml_adds_value.htm;
 NewsML, http://www.newsml.org/;
Publishing Requirements for Industry Standard Metadata (PRISM), http://www.prismstandard.org/;
WebEQ, (formerly http://www.dessci.com/en/products/webeq/ since replaced by
 XMetal, http://www.corel.com/xmetal;
 XyEnterprise XPP,
 MathMonarch 5, http://www.monarchsuite.com;
 Inera's eXtyles, http://www.inera.com;
Math on the Web Status Report (all editions), http://www.dessci.com/en/reference/webmath/status/;