Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.
@Jonas: Right click → Show source.
@Anthony: I can't reproduce the blue underlines. Are they always there, or only when you mouse over the titles? When you say the TeX in titles on math.SE is more nicely rendered, do you mean that the fonts are actually different, or just the blue underline issue?
@Will: One other person told me that he was having trouble with an old version of firefox. I've added a "(re)process with jsMath" link in the sidebar, but this is very much a temporary solution.
I need to do it twice, as I have for a few months anyway.
Hmmm. My first thought was that the main jsMath file was taking a long time to load, so the 1 second delay I built in wasn't enough, but if you had to hit the old reprocess link twice, I have no idea what's going on.
@Anthony: It must have been a caching issue. I haven't changed anything. I'm glad to hear it's better now.
Filipe, if you right click on the formulae, you can set the "zoom trigger" to "hover" and "zoom factor" to something like "400%", then the math texts will appear magnified when you mouse over. (Then again, recalling that the IPod has this new fangled touch screen technology, I am not so sure whether what I just described is even feasible on it.)
Also, does the math font not scale up the same as the text font?
Could it be a font/browser issue? I don't have access to a Mac to test at the moment. On Firefox/Linux, the $\mathrm{e}$-height and the "e"-height in normal text are the same. If anything I've found that the MathJax rendering is a bit wider (not sure about taller) than jsMath: a few equations that I edited to be less than the column width in jsMath now spills over.
In anycase, too bad you cannot right click, because there's also the option (after right-clicking) to "scale all math" by a fixed percentage relative to the text. Unfortunately it seems that MathJax stores the settings in cookies, so I don't see an easy way for you to make a global setting easily without going through the menu.
I confirmed the ipod problem this morning. It looks like math.SE has the same problem, but some inline math on mathjax.org looks okay, so maybe there's an easy way to fix it. I'll experiment a bit more tonight if I can get my hands on my wife's ipod. If somebody with an ipod confirms that this is a common problem on sites using MathJax, please post about it on the MathJax help forum
Not a bug, but a comment on what Michael Hardy reports: this definitely depends on font/browser/system.
My work computer runs Firefox 3.5 over Scientific Linux, and as I reported above, the 'e' heights between mathmode and textmode are almost exactly the same, none of these "bigger, bolder" problems people seem to be having.
Now that I am on my home computer, running Firefox 3.6.4 over Gentoo Linux, I found that the default setting leaves the inline math noticeably bigger than the surrounding text. I had to change the setting to scale all math down to 88% to improve things.
(Before the switch to MathJax, I am pretty sure the two machines rendered the site in the same way using jsMath [no size discrepancy]. But seeing that I also just updated Firefox yesterday, I am not sure how much of this artefact is due to MathJax, and how much due to the new version of Firefox.)
I am tempted to say that, insofar as size of the rendering goes, it may not be possible to please everybody.
The “ipod” problem is quite clearly a problem on the iphone as well.
Not only can you not “right click” on the idevice; you can't hover either, so you would have to set the maginfication trigger to click.
It would probably be better to set a suitable magnification for math, though. I somewhat naïvely thought I could set math magnification to 150% on the laptop, then hoping that this setting would become associated with my account and magically become active on the iphone. But that seems not to be the case. All of which indicates that another way to open the math settings menu would be most welcome.
@Harald: all the settings for MathJax are stored in cookies, which is why your attempt failed.
Is it perhaps possible to just have two buttons on the side bar that triggers MathJax to change the current math scaling percentage by +/- 5%? This presumably requires some javascript-fu that I don't have. This may temporarily alleviate the problem for those underprivileged folks without right clicks. The fact that things are stored in cookies means that essentially you just have to set it once for each device you own, so a slightly clunky, incremental interface should not be too much of a hardship for the users.
@Voloch: For right-click on the iPod, try clicking with two fingers at the same time. I haven't tried it myself, but this is something Mac is adopting at least for their laptop trackpads.
@Anthony: I used to have that with jsMath, but got used to it.
@Anton: I find that the default MathJax is rendering much too big on my regular old laptop (Firefox on Mac). I can right-click the correct the zoom factor, although the options are too many, and it takes some experimentation to figure out which of the many menus is the right one (Settings -> Zoom Factor does nothing, whereas Settings -> Scale All Math works). But actually I can also set it to MathML and be reasonably happy (again, Format -> MathML does nothing, whereas Settings -> Math Rendered -> MathML works). I think that eventually (or maybe it is already there) we should have a link to a how-to page on configuring MathJax. In any case, I like that I can chose whether to see it in MathML or CSS, and I also like that on Lynx everything just appears as raw TeX, which is easy enough to read.
(I was recently talking to a math professor at University of Oregon, who complained that MO was completely unusable. It turns out that for perceived security reasons, he refuses to use javascript. I find MO in lynx usable but not excellent --- better than, say, the New York Times, worse than Google Search, about on par with gmail.)
@theo: Right now the scaling of math is set to a default of 100%. For me, it looks too big in Chrome, but just right in Firefox, so I don't want to monkey with the defaults. It seems like inconsistent font size is a common issue.
Format -> MathML makes so that you can get MathML source from Show Source.
For right-click on the iPod, try clicking with two fingers at the same time.
Uh, how would the idevice determine the click location? Remember, the idevices have no analogue of the mouse cursor.
idon'tknow
It seems the only viable solution for touchscreen devices is that we try to insert buttons for "scale up maths" and "scale down maths" somewhere on the sidebar.
Would this be too annoying or take up too much space? If so, we could possibly do this only for certain browsers, but that may be unnecessarily complicated.
We should probably also bump this up to the MathJax developer.
That is a bit odd. It looks like your version of xulrunner is fairly recent (1.9.0.x), so it probably shouldn't cause too much difficulty. I don't think the Iceweasel version should matter, but maybe you can try to upgrade to 3.5 or 3.6 branch?
Also, I seem to remember the MathJax setup uses CSS to load fonts on the fly, I wonder in your case whether the bottleneck is at the network level (the loading of math fonts from the internet). Maybe use something like bwmon to check whether your network is saturated?
Here I am on 1.9.2.x with firefox 3.6 over Gentoo and the entire test page that you linked to loaded within 15 seconds.
I just had a go and while it wasn't as slow as yours, it was a bit slow (about 6-7 seconds for the first to appear, the others were slightly faster). So it may well be bandwidth as we're both in the same continent. If it is the downloading of the fonts then one possibility is to install the STIX fonts and change your MathJaX preferences to render as MathML. Then you get better looking output and faster as it doesn't have to download the fonts each time.
(You have to set your preferences anew on each site, I think)
Any chance we could switch back to jsMath until some more quirks are worked out? Preview is completely and totally broken in Opera, and I understand that there are significant issues in Chrome as well.
I'm very happy with MathJax, largely because I like MathML rendering better. (@Harry Gindi: there is the "reprocess math with jsMath" --- then again, it doesn't seem to do anything on my computer.)
I have noticed two small bugs with the MathML renderer (I'm using Firefox on Mac). One is that in-line equations will break where they shouldn't --- TeX knows a lot about what symbols it can break next to and which ones it really shouldn't, and I wish MathML / MathJax did too. The other is that MathML in-line equations seem to make the spacing much too big around lines, so that especially if I have math on consecutive lines, it looks strangely double spaced. Part of the problem might be that I have "scale all math" set to 110%, but that extra 10% is not explaining as big a gap as I'm seeing.
@Anton, if you can't reproduce this and want to see it, I can find you on Monday or something.
While writing this answer, I found that $0<\alpha<n/p-1$ rendered as “0<α0”. Inserting a space, as in $0<\alpha< n/p-1$, cured it. Is it a MathJax/markdown interaction phenomenon? (The moment I asked this, I realized I could test it by placing backticks around the formula. And indeed, with backticks it works fine without the extra space.)
The '<n' is getting interpreted as the start of an HTML tag and so is being sanitised by the SE software. Putting the space in breaks the match (since tags must start '<[a-z]'), putting it in code also breaks the match since the '<' gets converted to an entity before being processed.
(This is the sort of thing one should expect when running two formatters that don't integrate. I know that there's not a lot that can be done about that, but that doesn't stop me living in hope.)
(Seeing it written down is what gave me the clue, Harald. When you told me of this, I didn't visualise it so didn't spot it!)
In relation to Gil Kalai's question http://mathoverflow.net/questions/40561/a-combinatorial-abstraction-for-the-polynomial-hirsch-conjecture I found that "less than" symbols can be made to work if you put curly brackets around them like so {<}.
jc, that's because that also doesn't match the XHTML tag pattern. So any "NOP" after the < will do (but a } is only a NOP if there's a matching { before)
In TeX there is a big difference between < and {<}: The former is a relation, the latter is an ordinary math symbol and thus lacks the spacing that would surround a relation symbol.
I think there is a bug with the MathML rendering of operator names from MathJax. When I write "$\pmod 2$", for example, I should see something approximating "(mod 2)", whereas MathML (but not, it memory serves, the Javascript renderer) prints it as closer to "(m o d 2)". What's going on, I think, is that MathJax is incorrectly coded to conflate the commands "\rm" and "\operatorname". Now, the correct behavior for "$\rm{mod} 2$" is to print "m o d 2" (let's assume you've already coded the parentheses): the space in "$\rm{mod} 2$" is spurious, and clearly I am asking for the product of four terms, "m", "o", "d", and "2", and I am asking that the m,o,d be printed in roman. Operatornames, on the other hand, should be printed as words, not products. (I'll note that TeX does not add small spaces between letters in a product --- you only notice that it thinks about products and words differently by observing the handling of ligatures.)
Pages that have a lot of TeX run incredibly slowly for me when I use IE 8.
@Qiaochu: I was going to quip “don't do that, then”, but then I thought this might start a flamefest, so I shall restrain myself.