MathJax 1.0 is due to be released very soon (in a week or so?). I think there are good reasons for switching from jsMath to MathJax when that happens. MathJax seems to be more stable and have more uniform behavior across browsers. Also, it allows for the option of converting to MathML in browsers that can handle it.

I've implemented MathJax on faketestsite and I invite you to go try it out.

• It's configured to try to convert to MathML, but to use HTML-CSS if the browser doesn't support MathML (what browsers aside from Firefox support MML?).
• All macros that work on MO should work on faketestsite.
• The live preview should work.
• Math in the comments should work properly.

One issue I've encountered is that when I use Firefox and get MathML output, it looks like I don't have enough fonts installed. For example, \mathfrak and \mathcal both come out as upright roman font. Does anybody know how to solve this?

It's likely to feel a bit slow right now. There are two reasons for this. One is that I'm hosting MathJax on my home computer, so fetching the scripts and fonts may be slow. The other is that MathJax is still being optimized for speed. Both of these will be resolved by the time we start using MathJax on MO.

Edit: It must be an issue with mathjax (I just tested it myself), since the STIX fonts work fine on the nLab with fraktur.

hmmm ...

Also, by the way, is there any way for us to turn of mathml serving? Opera claims to support MML, but the support is awful (and it looks better the other way).

I think the answer is currently no, but I did email the MathJax team to tell them that I'm really looking forward to the addition of a control panel (similar to the jsMath control panel) which would allow users to set such preferences.

Okay, I installed Google Chrome with the express purpose of testing out the non-mathml rendering.

A.) Fraktur renders fine

B.) It is just as fast if not faster than mathml rendering (significantly faster than jsMath)

C.) It looks better

Proposal: Anton, could you totally disable mathml rendering until the bugs are worked out, then just put mathJax on the main MO site? The output is faster and nicer (even non-mathml) than the jsMath output.

I've thrown the iTeX torture test at it and I think I know why the fonts aren't rendering correctly. More details over at the fake test site.

No, I've been testing with firefox and chrome. Testing with opera wouldn't really give you important information!

More bugs!

Using Andrew's giant list of symbols (granted, many of them didn't render at all), I saw that all of the oversized symbols (direct sum, product, bigcup, bigcap, all integrals) are getting mangled in chrome (which is not receiving mathml from the software).

If someone has access to a macintosh computer, would you mind checking to see if the \prod gets sunk below the bottom of the line at this post?

Can somebody say some more about \prod's please? They render below the baseline for me, but they look correct:
: produced by MathJax in Chrome
: produced by jsMath in Chrome
: produced by LaTeX

@Andrew: can you explain exactly what problem the CSS you posted solves? Don't bold and italics work just fine already? I tried adding the CSS VA listed in this answer:

[mathvariant="double-struck"] {font-family: MathJax_AMS; } 
[mathvariant="script"] {font-family: MathJax_Script; } 
[mathvariant="fraktur"] {font-family: MathJax_Fraktur;} 
[mathvariant="-tex-caligraphic"] {font-family: MathJax_Caligraphic; } 
[mathvariant="bold-script"] {font-family: MathJax_Script; font-weight: bold;} 
[mathvariant="bold-fraktur"] {font-family: MathJax_Fraktur; font-weight: bold;} 
[mathvariant="monospace"] {font-family: monospace} 

it looks like it's supposed to get things like fraktur working, but it looks the same for me (yes, I cleared my cache). Does MathML mathfrak look right for anybody right now? Changing my font.mathfont-family variable in about:config didn't help.

Edit: there's a discussion about this on the MathJax Help forum, but as far as I can tell, it hasn't really been resolved.

Bold in mathematics produced by itex uses the mathvariant to render as bold but it doesn't display correctly on firefox without the CSS tweaks. There's a mozilla bug number (which is in the CSS that I posted) that deals with it. As this is something I just imported from the nLab to the nForum, I don't actually know much about it beyond that.

Regarding the fonts and VA's CSS, that would only work if you have the fonts already downloaded. Does MathJax do the same font stuff as jsMath? If so, it might not know that it needs to send MathJax_XYZ because the need is expressed in the CSS and not in what MathJax looks at. Try installing the fonts on your computer and having another look.

If that works, you could try using the @font-face directive to make the fonts available to browsers that support that, or simply say that everyone has to download these fonts to use MO!

I'm using the latest (5.0.375.86, as of this writing) version of Google Chrome on the latest version (10.6.4, as of this writing) of Mac OS X and I don't get any of the maths in http://faketestsite.stackexchange.com/questions/90/stress-test-using-mathjax/91#91 to render! All I see is the latex code. What am I missing?

When we were first setting up jsMath we had this problem too, and solved it by arbitrarily permuting the order we loaded various jsMath libraries in.

I think it was basically that we had to introduce some delay between when we loaded something and when we called it, because it would try to load asynchronously and end up stepping on its own toes. In this case there's only one script being loaded, and it's already being loaded after all the content is up, so I don't know what to do about it. As a hack, I could try something like setting a timer which attempts to process the page again after half a second.

I could also try messing around with this test page, which never seems to have that problem. But the math not processing happens pretty rarely for me. I don't have a way to intentionally reproduce it, which makes debugging tough.

I don't get it. Everybody seems to say \prod should be aligned on the baseline. But TeX surely centers it (and \sum too), and this looks right to me too.

@Harry: Yes, your pictures clearly show a problem, but Anton posted pictures just before yours, and I thought he was complaining about those. (Or did I misread his post? It's been edited, I don't know what it said when I responded.)

@VA: thanks for your detailed feedback. I tried explicitly loading those extensions, but I don't see any difference. In non-MML browsers, the fonts display correctly (as they did before), and in MML browsers, they still don't display correctly.

MathJax v1.0 is supposed to be optimized for speed. If it's still slower than jsMath, then I agree we should wait to switch.

If MML support in browsers is really as dodgy as you've described, then we should certainly use HTML-CSS by default. I'll look into the possibility of setting a preference cookie to use MML ... I'd like Andrew Stacey to be happy.

I got an email from Davide Cervone (the hero behind jsMath and, as I understand it, the driving force behind MathJax). He says that after MathJax 1.0 is out, he's willing to help us better integrate MathJax into MO. Woohoo! Between bugs being fixed in the latest svn updates and his help with implementation, I expect we'll be able to iron out most problems.

(Edit: corrected my misspelling of Andrew's name ... sorry.)

Update

MathJax 1.0 is now out! I've been in touch with Davide, who has been extremely helpful. Faketestsite is now running MathJax "for real" ... if we were to switch now, that's how fast/slow and robust/buggy it would be. I think it looks very good, but I wanted to get in one last round of public testing/suggestions before pushing the changes to MO. Two things worth pointing out:

1. (on MathML) Faketestsite will render math with HTML-CSS by default, but MathJax now allows you to change this to MathML by setting a cookie. Just right click on some math (except in the live preview because of how I've implemented it) and go to Settings > Math Renderer > MathML.
2. (on speed) MathJax does not (yet) have an equation cache like jsMath does, so if the same equation shows up many times, it has to reprocess it every time. For browsing, it should be at least as fast as jsMath, but the live preview is slower than you might otherwise expect.

My guess is that if people have complaints, it will be about the live preview. We could also do the live preview in the same way that math.SE does it (you have to wait a few seconds for before any math renders), but I like the approach on faketestsite much better. It has MathJax work in the background on a hidden div, after which it copies the html into the actual preview div. The result is that the preview is less responsive than the current jsMath version, but I think it's quite usable. Since the html is copied from another div, you can't get the MathJax control panel by right clicking. It also messes up MathML for some reason. I've emailed Davide to ask if it's possible to fix that. I'll try to remember to post a banner at the top of the site if I'm tinkering.

Please head over to faketestsite and give it a try.