[snakeanus]

> Say what you want about MathML but as of now it's the best standard we have for putting math on the web

I would argue that this holds true for (La)TeX, most sites use JS implementations of it and show the raw (La)TeX code as a fallback (or in an image form) if the user has JS disabled.

[sulizilxia]

That approach works well in terms of compatibility, but my sense is that it's usually much slower than MathML, and with equation-heavy documents can be extremely slow.

I might be unaware of something, though.

MathML isn't perfect, but I much prefer it to anything else on the web at the moment. I'd almost rather see an updated MathML spec than something new, especially one js-based.

[dunham]

MathJaX is sluggish because it's translating stuff after the page loads (and it takes the target browser + window size into account, so it's hard to move it out of the browser).

KaTeX can work as a preprocessor (but has less coverage of special environments and macros), and looks very promising to me. (And as long as you keep the original text in an attribute or hidden element, you can always migrate to something else later.)

The fidelity of MathML that I've seen has been terrible. The only people I've seen make it work is Apple, and there it's only working for their proprietary iBooks files. They render parts of the equation in SVG on top of the MathML, and they've got a special XML file that's tweaking a bunch of MathML layout parameters. I've tried copy/pasting the MathML into an epub, and it looks completely different, even when viewed in the iBooks app.