Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
on my system presently when following this anchor
http://ncatlab.org/nlab/show/Čech+cohomology#AbelianCechCohomologyDefinition
the browser takes me to the top of another subsection, different form the one that this anchor refers to. However the correct subsection is being highlighted (shaded in grey). Something funny is going on. Is it just on my system? Or am I missing something?
It works for me?
Takes me straight to the Definition section.
Zhen Lin, Todd, Thanks for the feedback! This means it must be something on my end. Sorry for the noise then.
I see now my problem. Something must have changed recently with the MathJax or whatever it is now that displays the formulas. In any case, displaying them takes much more time on my machine since recently. And the page reflows incrementally as the formulas eventually appear. In the case above, it took so long that I thought nothing was happening. But the missing reflow breaks the focus on the anchors.
Well anyway. It’s not a big issue. I just keep being amazed how much trouble it is in the 21st century to display a few symbols on a computer.
It’s not merely a matter of “displaying” symbols but actually arranging them. Imagine how time-consuming it must have been to manually typeset just one page of mathematics!
Yeah, but come on, today the most primitive computer chips run Doom. That also involves arranging some bits.
That’s not quite true. Most computers nowadays have dedicated hardware to deal with 3D graphics, and that’s quite sophisticated.
Anyway, as you observed, the real problem is that the page is reflowed as the mathematics is typeset. That is highly inefficient. But that seems to be a problem with MathJax in particular – I should have mentioned I’m using Firefox, so it’s rendered as straight MathML for me.
Ditto firefox here.
I also bet that doing the reflowing with javascript is a lot slower than in some other language. One of the many problems with mathjax (vs mathml).
In the new version of MathJax, there is an option to render as “Fast HTML” (right click a formula, Math Settings -> Math Renderer -> Fast HTML). I’ve been using this and it actually makes pages load at a reasonable speed (in Chrome, I refuse to use Firefox), albeit with lower quality.
Most computers nowadays have dedicated hardware to deal with 3D graphics,
Doom runs on printers, on ATMs, etc. e.g. here. The running joke is that these days every toaster runs doom.
I am well aware that me, the user, could do various things to have my symbols be displayed less than sluggishly. For instance I could sit down and write a program myself that does it. I am just saying that I am amazed that I have to do that, in the 21st century.
In the new version of MathJax, there is an option to render as “Fast HTML” (right click a formula, Math Settings -> Math Renderer -> Fast HTML).
Ah, thanks!
Anyway, as you observed, the real problem is that the page is reflowed as the mathematics is typeset. That is highly inefficient. But that seems to be a problem with MathJax in particular
MathJax’s problem, as with any non-native software (MathML is native) that wants to precisely position characters is that in current browsers there is no way to determine the extent the characters will take up until after they are rendered. Right now this means that every math expression has to be rendered twice - the first time to measure it and the second time to precisely position it (and reposition everything relative to it.) Firefox and Chrome I think still (I haven’t checked recently) have open bugs about determining size of text made into DOM structure without having to first render it (normal MathJax uses a hidden place for first rendering), but this seems to be a hard problem that would involve restructuring a lot of software.
I guess “Fast HTML” makes a guess about text sizes to do the first render pass and display, then batches together all the DOM changes to reflect the true sizes so they all happen at once and each change doesn’t trigger a reflow.
EDIT: The above is just guess about the various MathJax display modes since I don’t use it. “Fast HTML” may not actually use a second pass but only displays the guessed version. See Stack Exchange: Mathjax 2.5 alpha for more precise info. Maybe I’ll fire up Chrome to see what options it gives.
not sure if anyone wants to hear it, but just for the record:
with the entry geometry of physics – homotopy types it’s now at the point that on Opera with its MathJax hack it takes something like 20 minutes to display the formulas of the entry, on my machine.
I’ll have to add some alerts to the top of such pages, I suppose, advising people that the page effectively won’t display with some (?) or most (?) browsers
It’s fine on mine (FF 36.0.1), virtually instantaneous. However, the page is very long. Any thoughts on breaking it up into subpages?
Thanks, Todd. Yes, Firefox doesn’t resort to MathJax, it’s the MathJax engine that gets stuck. I will have to put a note on the page that it effectively only (?) works in Firefox. That’s sad, because it’s not meant for me, but for my audience, and it makes things awkward if the technology doesn’t simply work.
And, yes, the page is long. It is already one sub-part of what would otherwise be an even longer page, though. I can break it up, but it means increasingly more and boring work on my part. One would think computers would be there precisely to take care of trivial tedious tasks such as displaying a moderate chunk of text.
urs #14
not sure if anyone wants to hear it, but just for .the record:
with the entry geometry of physics – homotopy types it’s now at the point that on Opera with its MathJax hack it takes something like 20 minutes to display the formulas of the entry, on my machine.
I’ll have to add some alerts to the top of such pages, I suppose, advising people that the page effectively won’t display with some (?) or most (?) browsers
Does the nLab still have severe MathJax problems or have recent upgrades fixed them?
I tried loading in Chrome geometry of physics – homotopy types and the display seemed pretty good though it appeared that Chrome was running one of my CPU’s cores full out for maybe 3 minutes after loading. That page’s history doesn’t seem to indicate that it has been made much smaller since Urs’s comment.
I’ve been looking into whether the new Mozilla browser Servo will provide any hooks so that MathJax layout can be computed in 1 pass not 2.
1 to 17 of 17