Not signed in (Sign In)

Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.

Migration: Stack Exchange 2.0, and what this means for MathOverflow.

Bottom of Page

1 to 50 of 71

  1. <
  2. 1
  3. 2
  4. >
  1.  

    Some of you may have already noticed that Fog Creek, the company behind Stack Exchange, has radically changed their business model for Stack Exchange. You can read all the details here, and watch the community reaction over at meta.SE.

    Very briefly:

    • Most of the current Stack Exchange sites will be closed down in 3 months time.
    • Active ones are guaranteed 12 months of hosting in their present state (i.e. without further software improvements).
    • The current Stack Exchange software is currently being merged with the Stack Overflow software (bringing many enhancements, e.g. comment notifications), and they're referring to this merge as "Stack Exchange 2.0".
    • New sites can only be created through a community process, described in the blog post above.
    • Existing active sites will be able to migrate to the new software, and it seems likely that exceptions to the usual community rules for 2.0 sites will be allowed.

    Before anyone gets into a panic, remember that Fog Creek likes us. Anton was asked for his comments on the new system before it was publicly announced: we're in the loop and can expect special treatment.

    It seems that if we can get some exceptions to the usual policy, we'll probably be okay. I think for now the two most important things to ask for are:

    • Anton remains our benevolent dictator.
    • We continue to have access to complete data dumps.

    The current proposal for Stack Exchange 2.0 is that everything will be run by committee, and the public will at best have moderator powers (as opposed to Anton's administrator powers). Hopefully Fog Creek will see sense on this (there's plenty of noise to this effect on meta.SE), but hopefully we can grandfather in an exception. Access to complete data dumps (as opposed to the publicly accessible data dumps we've been providing) should obviously be an inviolable requirement, so that we have the option of switching out to different software. The unfortunate thing here is that if we're exceptional in having access to the database dumps, they may not actually do us much good, as the alternatives (Shapado, OSQA) won't have any reason to track changes in the database format over time, when every other Stack Exchange 2.0 site is siloed.

    • CommentAuthorHarry Gindi
    • CommentTimeApr 19th 2010 edited
     

    I feel like there should be a way for users to nominate administrators/moderators who have access to full database dumps etc. MO has just the right amount of oversight, which is really nice.

    • CommentAuthorNoah Snyder
    • CommentTimeApr 19th 2010 edited
     
    In addition to Anton's excellent points, I'm a bit worried about what's going to happen when they inevitably want to start posting ads in a year or two. As a community of academics I think people will be somewhat unhappy about a bunch of ads appearing. I'd much prefer the "someone puts MO in their grant" funding to something like "amazon ads to possibly relevant textbooks appear on questions" funding.
  4.  

    What's the likelihood of continued NSF funding for MO? (Assuming funding continues to be necessary or useful.)

  5.  

    I haven't posted about this on meta because I didn't really have any news yet, but I've been in touch with Fog Creek. It seems like they're very keen on not breaking sites that are working really well, but there's not a whole lot of information about what "migrating to SE 2.0" is going to mean. Basically, it sounds like they're going to migrate sites under conditions that are likely to vary from case to case (see this post). We have lots of time (they've promised to run MO as is at least until April 2011), so I expect to be able to carefully negotiate what's going to happen and to be able to post here to get feedback before committing to anything. If we can't get terms we like, we can move to a clone.

    The most important bit of "terms we like" is that I can still access full dumps and retain control of the domain, so we can pick up and change platform if they go in a direction we don't like. But really I'd rather not change anything at all (in particular, I'd be pretty unhappy about allowing ads), and for all I know they'll be fine with that. I don't think they're out to control MO at all. Anyway, I'll certainly keep y'all posted when I get some real information.

    @Harry: They are proposing some way of electing moderators, but my impression is that they haven't thought about the details very carefully yet. When I talked with them on the phone, I explained how important and useful dictatorial and admin powers have been at MO. The guy's reaction was complete agreement. I asked about how they plan to reconcile the clash between "community ownership" and a "benevolent dictatorship" (which I think is probably the best way to run an SE site, with lots of community input), he didn't really give me an answer. He said he'd bring up the issue when they meet to figure out exactly how this sort of thing will work. It could be that he was holding back information, but I really got the impression that they just wanted to let people know as soon as possible that they're not planning on going forward with the old business model and they really haven't figured out how the new system will work.

  6.  

    @fgdorais: MO doesn't have NSF funding. So far, we haven't had to pay anything. The funds that were lined up for MO were from Ravi's research funding (which I think comes from Stanford, not the NSF). I figured that in the long term I'd probably apply for NSF funding for MO. I haven't looked into it carefully, but the impression I get from talking to people is that the NSF would jump at the opportunity to pour money into MO.

    • CommentAuthorHarry Gindi
    • CommentTimeApr 19th 2010 edited
     

    @Anton: I meant so we could elect you! I agree that the benevolent dictatorship model is the right one for MO.

  8.  
    I want to support Anton's impression: it seems very likely to me that the NSF and other government organizations would want to fund MO. Moreover, a lot of mathematicians that are PI's on NSF grants would probably jump at the chance to link their research activities with MO: it's a "broader impact" bonanza. Count me in on the people who are waiting to be told that they can try to divert grant money to MO.
  9.  

    Count me in on the people who are waiting to be told that they can try to divert grant money to MO.

    :-)

    • CommentAuthorAndrea
    • CommentTimeApr 20th 2010
     

    I read the whole announcement and find it a very sad situation. I think it is at best irresponsible to create a software and make agreements for payed subscriptions, and then shut it down before it even gets out of beta. Ok, they have a new model, but in this new model it seems they are owner of everything (domain name, private dumps...) which is not very promising.

  11.  

    I'd say that this is your golden opportunity. Now that you have shown that MO works, go to all those with money and say "Look, here's a nice bandwagon for you to jump on. We have 12 months to develop our own version of this system, with all the features that we want. Would you like to give us lots of money to do so?".

  12.  

    But Andrew, we don't have any developers! Moreover, I think it's unlikely that this would be a good use of mathematicians money, given that at this point we expect that StackExchange 2.0 will work out okay for us.

  13.  

    Scott beat me to it. MO is a big enough project. There's no reason to shovel more on our plates. I'll add that there are also existing open source alternatives that I'm keeping an eye on, like OSQA, Shapado, and Askbot. Not only am I not interested in reproducing the code, but these alternative platforms are already being quickly developed because so many SE 1.0 sites have to migrate or die in 3 months. The right thing for us to do if we don't want to be an SE 2.0 site is to wait a few months to see what happens with those platforms. One of them will likely come out on top.

    At the end of the day, it'd be nice to have somebody whose primary job is to do things like maintain code, optimize server setup, implement clever caching, fix bugs we report, and implement features we request. Starting our own project and hiring somebody like this is way more headache and more expensive than I'd like. Doing it myself is out of the question: that's not what I want my job to be. If the SE team is willing to do it on terms we like, great! I know they're really good at this stuff. If not, there are other people willing to fill that role.

  14.  

    I just want to state that I would hate to see MO get ads. I feel guilty asking for this when I am not contributing any funds to the site, but any company which I can imagine advertising on MO would lower its class. If it comes down to that, I think we should find out how much SE would need us to pay them to get ad-free hosting.

    • CommentAuthorRegenbogen
    • CommentTimeApr 20th 2010
     

    My god, the screenshots at OSQA, Shapado and Askbot look much like StackExchange. Why not adopt one of these in a short while? If there is not much difference between a commercial software and a public domain software, then we being academicians and non-profit guys, we should prefer the public domain software.

    • CommentAuthorKevin Lin
    • CommentTimeApr 20th 2010 edited
     

    Approximately how much money does MO cost per year?

    I'm guessing that it would not be hard to raise at least a couple thousand dollars a year by asking for donations from users. I would certainly donate.

  17.  

    Somewhere between 1300 and 1600 a year. I'm pretty sure that money isn't a problem at the moment though.

  18.  

    @Regenbogen: Looks is not the same as functionality. Important things are still missing. For example, you specifically might be interested to know that none of those platforms currently have a way to post as an unregistered user, making it much more of a hassle to post anonymously (both for the person posting, and for moderators watching out for vote fraud). It would also be an unacceptable barrier to entry: lots of mathematicians use MO for a long while before registering.

    @Kevin: It hasn't cost anything yet. The "entry level" SE price plan was $1560/yr, but MO now gets well over 10^6 views per month, so we would be in the $4800/yr plan. If we end up under that plan, money won't be a problem. It wouldn't come to collecting donations from individual users for a long while.

    • CommentAuthorHarry Gindi
    • CommentTimeApr 20th 2010 edited
     

    It wouldn't come to collecting donations from individual users for a long while.

    Especially given the number of very generous offers of funding from people like Pete and the funding from our great benefactor.

  20.  

    I would say that there's one humongous reason for switching to an open source system ASAP: rendering of mathematics. Just imagine if you could get it right! Yet again, I've opened a seemingly interesting question just to find it rendering as gibberish (okay, slight exaggeration) and so not worth trying to decode to see if the question really was interesting.

    • CommentAuthorAndrea
    • CommentTimeApr 21st 2010
     

    Am I right in understanding that we do not have access to some data dumps now? For instance I think we do not have the data of the users. So the problem of migration is not really some feature missing, as these projects are open source and we may fork them and add what we need (even if I think they will develop quick enough), but rather that we are more or less tied to the StackExchange guys. Which is especially annoying, since we do not know there business plans, and the only guess we can make is that they will want to put ads.

  22.  

    @Andrew: I like jsmath better than iTeX =p.

  23.  

    @Harry, you're missing the point. I don't care about the input. I care about the output and about the fact that it takes so long because it is computed by the client, not the server.

    @Andrea: I think that Anton has the full data dumps. For some reason, he doesn't want to share it all with us.

    @Anton: I would bet that if you emailed the people at OSQA and said, "I'm pondering migrating MathOverflow to your system. Here's why I can't do that yet ..." then you'd find that all these problems would silently and softly vanish away.

  24.  
  25.  

    @Andrew: I mean that jsMath renders the same on all computers and all browsers. MathML rendering is at best spotty.

  26.  

    All I can say to that is:

    lynx www.mathoverflow.net

    Seriously, javascript is not the same across all browsers and all computers. One would like it to be, of course, but there's no standard for javascript in the same way as there is for MathML. A case comes to mind where the same browser on the same computer produced different results for the same page when viewed as two different doctypes.

  27.  

    @Anton: I would bet that if you emailed the people at OSQA and said, "I'm pondering migrating MathOverflow to your system. Here's why I can't do that yet ..." then you'd find that all these problems would silently and softly vanish away.

    Actually Scott and I were having that sort of conversation with the people at OSQA several weeks ago, but somehow nothing became of it. I've been in touch with them more recently, and it sounds like things are in better shape now for us to play around with OSQA. I'll keep you posted if there are developments in this direction.

    I'm certainly interested in alternative platforms, but I think calling for MO to switch now is irresponsible. There are lots of things that can go wrong, even if we ignore features. None of the alternative platforms have run a site as large as MO; there could be scaling issues. If MO is going to switch, I want to be sure that (1) the transition will be smooth, and (2) there's a good reason to do it. Right now, I believe that neither criterion is satisfied.

    I think that Anton has the full data dumps. For some reason, he doesn't want to share it all with us.

    I have access to full database dumps. I don't share them because they contain private data (like email addresses, IPs, and exact voting histories). I try to put as much into the sanitized dumps as I can without compromising anybody's privacy. If there's something you think I should include in the public dumps which I'm not including, suggest it on the public dump thread.

    All I can say to that is:
    lynx www.mathoverflow.net Seriously, javascript is not the same across all browsers and all computers. One would like it to be, of course, but there's no standard for javascript in the same way as there is for MathML. A case comes to mind where the same browser on the same computer produced different results for the same page when viewed as two different doctypes.

    I don't follow. The math on MO looks awesome on lynx, but sites using MathML look awful. I have yet to be convinced that the "jsMath model" isn't the right one. MathJax is going to be much more robust and faster than jsMath. I see no reason not to use MathJax regardless of platform.

    • CommentAuthorAndrea
    • CommentTimeApr 22nd 2010
     

    The best way to render math today would be to use mathml, and then check with a javascript library if support for mathml exists; if this is not the case use javascript rendering. In this way

    1) Users without javascript will use mathml

    2) Users without mathml but with javascript will use javascript

    3) Users with mathml and javascript will use mathml

    A similar approach is what is generally used for new HTML5 and CSS3 features, which are not fully supported everywhere.

  29.  

    I'm certainly interested in alternative platforms, but I think calling for MO to switch now is irresponsible.

    If you read what I originally wrote, you'll see that I wasn't calling for MO to switch now but to do exactly what you are doing.

    I should also "declare an interest" in that I'm interested in this not really for MO but because having used it, I think it would be fantastic to have as a aid for a course. But that would be too small to fit the SE model, and so - quite apart from my ideals on this matter - FOSS is the only way to go.

    @Andrea: that's almost exactly what I think except that I see no reason to use javascript. If a browser can't cope with mathml, then if it can take images, send a picture of the maths, and if it can't, send the original source as the text. I think that that's the most accessible way to do it.

    @Anton: do you have some fancy configuration of lynx that can handle javascript? When I look at an MO page using Lynx then I just get the LaTeX-ish source. But then I don't actually use Lynx all that much so maybe I haven't set it up correctly.

    • CommentAuthorAndrea
    • CommentTimeApr 22nd 2010
     

    By the way (excuse me for the OT): is there any reason you use Lynx at all to browse MathOverflow? When I develop, I usually try to be careful (if possible) that no content gets lost when visiting a site with a text browser, but I wonder how many people still use, and how often.

  31.  

    I thought lynx was dead, and that w3m had taken its place these days. But anyway, one reason to check with a text-only browser is that it might give you a rough indication of accessibility for blind users.

  32.  

    I guess that this is an answer to the question of opening up a new MO for undergrads though! Now people won't come here to complain!

  33.  

    @Harald: very, very rough. Accessible technology is quite sophisticated these days and can read quite complicated documents. But they need to be standards compliant first, hence MathML.

    @Harry: if you're referring to my comment, then only my students will be allowed access. But if you looked at SE, you'll see that there's a proposal for a more general mathematical site than MO, so the "riff-raff" will have somewhere to go once that gets off the ground.

  34.  

    But if you looked at SE, you'll see that there's a proposal for a more general mathematical site than MO, so the "riff-raff" will have somewhere to go once that gets off the ground.

    Yes, this is the same thing that I was talking about.

    • CommentAuthorjonas
    • CommentTimeApr 23rd 2010 edited
     
    Unlike some of you, I don't think ads would bother me. (I'm thinking of tasteful ones, not the annoying colorful popup kind of ads.) If I want the site to run, I also want them to gain the money from ads. I just mentally filter out the ads from webpages I look at.

    Update, clarifying: this only applies to community sites administered by volunteers, where I'm benefitting from a free service.
  36.  

    do you have some fancy configuration of lynx that can handle javascript?

    No. I see LaTeX when I view MO in lynx. That's the best thing to see if you don't have MathML support or javascript (try reading a MathML page in lynx!). I guess I agree that if you can view MathML, that's better than jsMath's html+css display. Incidentally, I discovered today that MathJax can produce MathML output if your browser can handle it and html+css output otherwise (the MathML output is supposedly generated much faster). See the "MathML Output" section near the bottom of the beta 2 documentation. This is slightly suboptimal because if you can handle MathML, but have javascript disabled, you get LaTeX. However, I think this is much better than getting MathML when your browser can't handle it.

  37.  

    I've put up a MathJax test page with the MathML→html+css→plain LaTeX preference order. Feel free to give it a try. When I try it, the html+css version is pretty slow, but the MathML version is pretty fast (maybe even faster than the jsMath version of the test page).

  38.  

    Is there any way to tell if it's mathml, latex, or the thing that's like jsMath?

  39.  

    I was checking using the "inspect element" tool in chrome and the corresponding firebug tool in firefox. MathML looks like

    <mtd><mi mathvariant="italic">σ</mi><mo stretchy="false">(</mo><mi>y</mi><mo stretchy="false">−</mo><mi>x</mi><mo stretchy="false">)</mo></mtd>

    html+css looks like

    <span style="position: absolute; clip: rect(0.16em 1000em 1.526em -0.516em); top: -2.293em; left: 50%; margin-left: -1.731em; "><span class="mtd" id="MathJax-Span-30"><span class="mrow" id="MathJax-Span-31"><span class="mi" id="MathJax-Span-32" style="font-family: MathJax_Math; font-style: italic; ">σ</span><span class="mo" id="MathJax-Span-33" style="font-family: MathJax_Main; ">(</span><span class="mi" id="MathJax-Span-34" style="font-family: MathJax_Math; font-style: italic; ">y</span><span class="mo" id="MathJax-Span-35" style="font-family: MathJax_Main; padding-left: 0.222em; ">−</span><span class="mi" id="MathJax-Span-36" style="font-family: MathJax_Math; font-style: italic; padding-left: 0.222em; ">x</span><span class="mo" id="MathJax-Span-37" style="font-family: MathJax_Main; ">)</span></span></span></span>

    LaTeX looks like

    $\sigma (y-x)$

    • CommentAuthorAndrea
    • CommentTimeApr 26th 2010
     

    How do I disable MathML to test the speed for html+css?

  41.  

    @Andrea: I don't know. For me, MathML works in Firefox, but not in chrome. Hopefully somebody can give a proper answer to that question.

    • CommentAuthorAndrea
    • CommentTimeApr 26th 2010
     

    Thank you. I tried the MathJax rendering in Chrome and it was indeed hell slow. Not only that, but the rendering was wrong: many symbols were misplaced (bad vertical alignment) and some symbols were only half-displayed.

    Are there any plans for jsMath to output MathML if possible? That would be close to ideal.

  43.  

    Hey, Anton, is there any way you could take some screenshots of what the different renderings look like?

  44.  

    @Andrea: yeah, I'm having the same problem where various symbols (especially parentheses) have crazy vertical alignment. I get the same sort of problem on the MathJax preview page, but only at certain magnifications and now quite as crazy. I'll file a bug report when I get home (the internet in my office is being painfully slow right now).

    Are there any plans for jsMath to output MathML if possible? That would be close to ideal.

    jsMath is no longer being developed. MathJax is the way of the future: it already outputs MathML, and once it is optimized, it should output html+css as fast as jsMath does.

    • CommentAuthorAndrea
    • CommentTimeApr 26th 2010 edited
     

    I see the MathJax preview page correctly with Chrome. And their first FAQ states:

    The symbols are misaligned in beta 2. What happened?

    The fonts changed in beta 2. You must completely replace your beta 1 web fonts installation with the beta 2 web fonts. Both the actual font files and the font metric information in the MathJax code changed in beta 2. If you only update one or the other, they will be out of sync, and characters will appear badly misaligned. Remember to unzip MathJax-webfonts-beta2.zip! Merely replacing the zip file will not update your installation unless you unzip it. This also applies to svn users, who will need to delete the MathJax/fonts directory and unpack the new fonts.zip file. If you get the latest version via svn, the fonts.zip file will take some time to downlaod, as it is approximately 13MB in size, so be patient.

  46.  

    @Harry, re: screenshots. Anton is a busy man. :-)

  47.  

    @Andrea: I saw that, but I'm using the latest svn fonts and code. If I just use MathJax beta 2 and the corresponding fonts, it works right: test page. It must be that something isn't up to date in the public svn repo.

    Unfortunately, I just took a couple of screenshots, undermining my image as a busy man ... I have to learn to delegate these things. Anyway, HTML-CSS:
    html-css
    MML:
    MML

  48.  

    HTML+CSS still looks significantly better.

  49.  

    @Harry: From the beta 2 documentation:

    At some point, MathJax may include a control panel that will allow users to select preferences on their own, but that is not available currently.

    You may also be able to use different fonts with MML, but I don't really know anything about how that would work.

  50.  

    I got an email from the SE team which clarified some things for me. At some point, the SO code base was forked and SE 1.0 was born. The two projects developed pretty much independently, occasionally swapping code, but basically diverging. Now the plan is to merge the projects. Because of how they want SE 2.0 to work, it makes more sense to use the SO code base as the starting point and merge features from SE 1.0 into it rather than the other way around. This (mostly) explains why SE 1.0 sites won't get new features until they migrate and why it's taking so long before we hear anything about what migration will mean.

1 to 50 of 71

  1. <
  2. 1
  3. 2
  4. >
Back to Discussions Top of Page