Science After Sunclipse

I often joke that the technology for typesetting equations on the Web is of all-consuming interest to about twelve people. This is, of course, the kind of joke I make out of bitterness: it's a joyous day indeed when I meet one of the other eleven! Still, things can't be as bad as all that, since Brian Hayes managed to convince American Scientist — one of the few science magazines which hasn't yet seriously pissed off working scientists — to print an article on the problem. This bit of the introduction lays out the situation fairly well:

The truth is, the basic protocols of the Web offer almost no support for rendering mathematics or other specialized notations such as chemical formulas. Presenting such material on a Web page often requires software add-ons or plug-ins to be installed by the author or the reader or both. Fine-tuning the display of mathematics can be a fussy and finicky process, not much easier than formatting equations with a typewriter. The results sometimes render differently—or not at all—in various Web browsers. This is a sad situation: As the Web has evolved into a thriving marketplace and playground, the scholarly and scientific community that created the technology has not been well served.

The confused state of online mathematical typography is worrisome as well as sad. In years to come the Web will surely be the most important conduit for scientific information. Already it is a major channel for distributing publications and preprints in many disciplines, and it is becoming a venue for less-formal jottings and conversations—everything from homework assignments to blogs. Ideally, the Web would serve as an extension of the blackboard where people gather to talk about science and math during coffee breaks. We need chalk for that blackboard.

The problem is not one of simple neglect. Over the years there have been many earnest efforts to build a reliable facility for writing and reading mathematics online. The trouble is, no one solution has yet gained the kind of widespread adoption that would make it a standard, supported in mainstream Web servers and browsers.

You took the words right out of my hacked-together LaTeX-rendering plugin!

Hayes goes on to describe the various technologies cooked up to deliver mathematics over the Web: PDF, MathML, assorted ways of rendering TeX into images. Each has advantages and drawbacks. Pre-rendering equations written in TeX (or some flavour thereof) as GIF or PNG images is how I write things like this:

Or this:

Most content-distribution platforms require an extra plugin to insert equation images. For blogging software like WordPress and Movable Type, there is LaTeXRender, which takes a string of text like \frac{x^2}{a^2} + \frac{y^2}{b^2} = 1 and spits out an image. WordPress.org is still fixing bugs on the implementation which they use for the blogs they host, but glitches aside, it seems to function pretty well. The Movable Type analogue requires MT 4.0 or higher, and has not yet been installed on ScienceBlogs.

(How do I make the equations I use? Magic. Laborious, laborious magic.)

If you display your equations as pictures, they'll work in just about any browser, as the typesetting is done server-side ahead of time (rendering plugins typically provide the TeX source as alt text for accessibility purposes). This is convenient for people who are just Googling around for homework help and don't have the time or the inclination to install special browser plugins. Wikipedia does mathematics in this way, for example. One downside is that because the images are computed beforehand, there's no way to make them look right to the end user: browsers can be configured and reconfigured in all sorts of ways, and so the colours and sizes of the equations can end up completely wrong for their surroundings. The problem only gets worse when people start reading your content through RSS feeds. . . .

At the opposite technical pole is MathML, which is what its name says, a Markup Language for Mathematics. Of this, Hayes says,

On the Web, however, MathML has not exactly swept away the competition. One reason is lack of support in browsers. In the early years, the only way to read MathML Web documents was with plug-in software. More recently a few browsers—notably those of the Mozilla family, such as Firefox—have gained native support for MathML. But there is still confusion over how MathML content should be embedded in an HTML document. Moreover, MathML has not solved the fonts problem; readers are still responsible for installing appropriate fonts.

Another factor inhibiting the spread of MathML is simply that TeX is deeply entrenched, particularly in physics, mathematics and computer science. If you live in a TeX-centric universe—I have a friend who even writes love letters in TeX—it's hard to see any benefit of a new and very different language.

Who's been peeking into my LaTeX sonnet files?!

If you're willing to make your readers go through a couple software-installation steps, you can use a platform which turns TeX into MathML which is then rendered by the reader's Web browser. This is the approach taken by The n-Category Café, for example, a blog which serves a sufficiently specialized community that resolving a few TeXnical issues is seen as a reasonable price to pay for mathematics which meshes well with its surroundings.

An intermediate solution is the JavaScript-based tool jsMath, which is basically a TeX interpreter stuffed inside a JavaScript program. This allows equations to be rendered on the fly, but it can get bogged down when lots of content has to be processed. The same caution goes for Josh John Forkosh's MathTeX, which is pretty easily integrable with everything.