tea.mathoverflow.net - Discussion Feed (The Sage Cell server on Mathoverflow)

jasongrout comments on "The Sage Cell server on Mathoverflow" (21765)

jasongrout — Mon, 01 Apr 2013 15:26:25 -0700

Peter just emailed me a link to this discussion. I started (and continue to maintain) the Sage cell server project. I've got to run, but here's a quick response to the three points raised:

(0) We are ramping up the infrastructure soon to multiple servers. We've been running it on a single (but pretty high-end) server. We're moving to 4 virtual servers for fault-tolerance and load balancing. However, if (and before) this became official on MO, I'd feel better asking William Stein (owner of the servers, and the guy who started Sage) if it's all right to use his computers (if that's what you want to do).

(1) You could always set up a server that proxies requests to the cell server. For example, you could set up HAProxy to transparently reverse proxy requests, and at MO just use something like sagecell.mathoverflow.net. If our server goes down, you could redirect to a different server. Or you could set up your own server, of course. You'll also have the sage code, so you could set up something so that if for some reason the sage cell server isn't available, the page could still render the code as plain text.

(2) On the sage cell server, a user could do something like this and thereby embed whatever javascript they wanted into the page. You could prevent that by sanitizing the html that comes back from the Sage server, of course.

Scott Morrison comments on "The Sage Cell server on Mathoverflow" (21764)

Scott Morrison — Mon, 01 Apr 2013 14:09:56 -0700

Re: (0) and (1), we could run our own server on an Amazon EC2 computer. This costs money [1], but not a huge amount, and it would be pretty cool.

(2) is indeed a big worry. Peter's reasons are reasons to worry less, but sticking user content in script tags is so terrifying you have to have more than a list of reasons not to worry. :-)

[1] (At the barest minimum, about $30/year, for an anemic processor that would be worthless to us. I'd guess for a few hundred a year we could run something usable, but I don't have a good way to estimate how much CPU a Sage Cell server really warrants.)

voloch comments on "The Sage Cell server on Mathoverflow" (21763)

voloch — Mon, 01 Apr 2013 12:11:03 -0700

@peter re (1). I am not worried about Sage folding but they could do something as simple as change the URL to break the scripts here. Relying on an outside resource is always brittle.

peter.krautzberger comments on "The Sage Cell server on Mathoverflow" (21762)

peter.krautzberger — Mon, 01 Apr 2013 12:04:17 -0700

re (0) very much so. The Sage Cell server is pretty slow as is. See my earlier comment.

re (1) Sage has been NSF supported for a while and is managed by an independent foundation.

re (2) Two reasons not to worry: a) Sage works server side so any code gets sent to Sage as pure text and is interpreted there and sent back. b) the text should be sanitized into pure text coming from the MO input anyway (not sure if sage would have problems with that)

Anton Geraschenko comments on "The Sage Cell server on Mathoverflow" (21761)

Anton Geraschenko — Mon, 01 Apr 2013 11:35:56 -0700

Based on a quick look, this seems doable, and it certainly would be cool. There are a few concerns that come to mind:
(0) Let's make sure we don't send a ton of traffic to some sage cell without it expecting us.
(1) Support. If the ground moved underneath us, it would be good to know that lots of posts won't become unreadable. I'm not too worried about this because not too many posts are likely to use the new functionality, and the ground isn't likely to move much.
(2) Security. Sticking user content into script tags raises a huge red flag. I'd like to be really sure that nobody could do something malicious. It's probably okay to trust the sage cell operators, but definitely not okay to trust everybody on the internet.

François G. Dorais comments on "The Sage Cell server on Mathoverflow" (21757)

François G. Dorais — Mon, 01 Apr 2013 10:08:19 -0700

That should work since that happens after the software strips bad html tags. It's worth a shot; I'll ping Anton and Scott.

peter.krautzberger comments on "The Sage Cell server on Mathoverflow" (21756)

peter.krautzberger — Mon, 01 Apr 2013 10:04:15 -0700

If you do it right, you wouldn't need a script tag in the post. I'm guessing you looked at the about page? That seems to be written more with raw html authoring in mind. Since you enter text via the MO input field, you'd just write something like \begin{sage}\end{sage} which then gets turned into script tags when inserted in the page.

François G. Dorais comments on "The Sage Cell server on Mathoverflow" (21752)

François G. Dorais — Mon, 01 Apr 2013 09:41:08 -0700

You can't use a script tag in posts. Is there another way?

peter.krautzberger comments on "The Sage Cell server on Mathoverflow" (21751)

peter.krautzberger — Mon, 01 Apr 2013 09:24:05 -0700

Here's a link to the Sage Cell server about page.

peter.krautzberger comments on "The Sage Cell server on Mathoverflow" (21750)

peter.krautzberger — Mon, 01 Apr 2013 09:18:33 -0700

Although I'm not active on MO, I think that's a great idea. You wouldn't need a separate Sage Cell Server, actually (that's the point of Cell really). However, a separate server might be preferred for speed and reliability.

Much like MathJax, Sage really only needs a javascript call, so no major changes to the SE software should be needed (though probably a few minor ones to integrate the output). In fact, you could probably write a MathJax extension to wrap it all in one.

isasdoes comments on "The Sage Cell server on Mathoverflow" (21746)

isasdoes — Mon, 01 Apr 2013 06:00:39 -0700

The Sage Cell Server enables embedding of live Sage computations directly into any web page. These computations can easily include 2d and 3d plots, sliders, buttons, etc
http://jointmathematicsmeetings.org/amsmtgs/2141_abstracts/1086-00-2238.pdf

Would it be a good idea to have the Sage Cell Server on Mathoverflow?

Could this be installled on Mathoverflow with some javascript or would it require Stackexhange to modify their software?

What existing questions would have benefited from including a live computation widget?

What sort of questions in general would this be good for?

Would it be used often?