Yesterday I found myself composing a reply to LW in one tab while using 2 other tabs to look up text and URLs so I could paste them into my reply. If I understand what you want correctly, MathJax would have run in all 3 of the tabs (even with Chrome's process-per-tab architecture) and the delays and pauses in things like scrolling would have had a chance to interrupt my train of thought.
Also, some of us have tried Chrome and after consideration, decided it is worse than what we are currently using. Google's motive in introducing Chrome was to neutralize the threat to its revenue stream posed by Microsoft's Bing and other Microsoft initiatives by undermining Office and desktop software sales in general, which is very different from my main motivation for using a web browser, which is to have a quick and reliable and non-glitchy and non-crash-prone way read words (and math symbols) written by many others and write words (and math symbols) that will be read by many others.
according to my informal benchmarks, the memory cost of loading some JavaScript code is actually very small
There is no upper bound on the memory cost or time cost of loading a file of JavaScript, as is true of any Turing-complete programming language. With this particular JavaScript file, I worry about time cost more than memory cost because I can imagine that it converts LaTeX to images in the browser and doing that conversion in an interpreted language seems time-intensive.
In summary, what you write leads me to believe that although a few LWers might like MathJax here, you do not really know whether it will have adverse effects on the majority of LWers. Note that these adverse effects might be diffuse (meaning they worsen the user experience of many people just a little) and difficult for the average LWer to attribute to MathJax -- just as it was difficult for me to attribute my frustration to Google Analytics's JavaScript until I made the experiment of preventing it from loading.
ADDED.
I would be interested in your answer to my question bottom of grandparent.
Moreover, I am curious as to the source or cause of your enthusiasm for LaTeX on Less Wrong. A lot of people here are in the habit of reading the output of LaTeX (printed on paper) since it is used for most math papers and many (most?) science papers. I suppose you are one of them; aren't you? Are you also in the habit of writing LaTeX?
Also, I should warn you that as a general rule, I tend not to continue threads that consists of just 2 people replying back and forth more than 3 reply-counterreply cycles deep unless the thread gets upvoted a lot. I mention this so that you do not write a long reply to this and then get disappointed when I do not reply or reply with just a summary of my position.
It looks like you've expanded your post after I replied, so here's an addendum.
Moreover, I am curious as to the source or cause of your enthusiasm for LaTeX on Less Wrong. A lot of people here are in the habit of reading the output of LaTeX (printed on paper) since it is used for most math papers and many (most?) science papers. I suppose you are one of them; aren't you? Are you also in the habit of writing LaTeX?
I'm used to reading and writing LaTeX, yes. The source of my enthusiasm for it comes in two parts:
Because occasionally I want to be able to
In the next month, the administrators of Less Wrong are going to sit down with a professional designer to tweak the site design. But before they do, now is your chance to make suggestions that will guide their redesign efforts.
How can we improve the Less Wrong user experience? What features aren’t working? What features don’t exist? What would you change about the layout, templates, images, navigation, comment nesting, post/comment editing, side-bars, RSS feeds, color schemes, etc? Do you have specific CSS or HTML changes you'd make to improve load time, SEO, or other valuable metrics?
The rules for this thread are:
BUT DON’T JUMP TO THE COMMENTS JUST YET: Take a few minutes to collect your thoughts and write down your own ideas before reading others’ suggestions. Less contamination = more unique ideas + better feature coverage!
Thanks for your help!