Summary

  • On 7 April 1998, W3C issued MathML 1.0, its first XML application to become a Recommendation, so mathematical notation and mathematical content could travel as Web data.
  • Presentation markup described the arrangement a reader sees; content markup described operations and relationships that software could process. The Recommendation treated usable tools and rendering as implementation work still to be done.

Imagine writing a paper on the Web and discovering that your equation has to pose for a portrait. The 1998 W3C MathML Recommendation described a common workaround: make the formula a GIF and place it beside the text. The page could display the portrait. It could not search inside it, select a term, copy a subexpression into another tool or ask the mathematics to speak. The people who invented a medium of linked, inspectable documents had handed their formulas a picture frame.

W3C's account of the problem was unusually physical. An image sized for one font might look tiny when the reader enlarged the page. A white background baked into its anti-aliased edge could leave a pale halo on a different page color. The baseline could sit in the wrong place. A screen-resolution formula printed badly next to sharp text. Even a conscientious alternative description could only describe the picture; it could not turn the pixels into a reusable expression. The GIF was an excellent postcard and a poor collaborator.

MathML 1.0 made a more ambitious bargain. It used XML to carry mathematical material as markup. One family of elements expressed presentation: fractions, superscripts, rows, operators and the other arrangements that tell an eye how to read notation. Another expressed content: the operations, operands and relationships a machine might search, transform or pass to a mathematical application. The Recommendation counted 28 presentation elements and 75 content elements. W3C was giving the equation a wardrobe and a skeleton, and inviting software to work with both.

Those two layers met under rules, rather than by throwing tags into a bag. The Recommendation's mixing chapter explains that a presentation expression still has to render sensibly and a content expression still has to carry a coherent meaning. It even shows an operator that needs an annotation to say what it means. A drawn mark alone may look familiar to a reader while leaving a program guessing. Structure improves the conversation only when the participants agree which part is appearance and which part is an instruction about mathematics.

On 7 April 1998 W3C announced MathML as its first XML application to reach Recommendation status. The specification expected equations to be generated chiefly by editors and converters, not by researchers hand-typing forests of tags. It set ambitions for high-quality display, printing, speech, braille, searching and exchange with mathematical software. The document was candid that good browser rendering and authoring depended on actual implementations. W3C also pointed to MathML support in its Amaya testbed. That is a useful bridge to the browser/editor story, while the larger MathML story lives in the equation itself.

The small victory was conceptual and durable: a Web equation could be a document participant. A reader could see notation; another tool could receive structured content; a future editor could work on a part instead of replacing a whole image. The equation left its museum wall and took a place at the workbench. Whether each browser and editor made good use of that invitation was a separate engineering contest.