I was going to download the data dump and answer this question myself, but I keep getting an unable to connect to tracker error.

It certainly seems like it though and it would make sense seeing as there are a lot of people running around tagging duplicates and making sure those questions get removed. The only questions left seem to be edge cases of interest to only a handful of developers.

It this going to be a problem? Will the community begin to dwindle because of it? How can it be fixed?

  • statoverflow.com – deleted Sep 6 '09 at 3:47
  • I haven't noticed any "slowdowns." Are you visiting SO more on Sundays? – mmx Sep 6 '09 at 3:48
  • I've been visiting it all week. It seems slower to me. If I had the data dump I could be sure. – Spencer Ruport Sep 6 '09 at 3:50
  • See also here for the data dump: tejp.de/files/so-export The current dump is the so-export-2009-08.2.7z file. – sth Sep 6 '09 at 5:01
  • If you looking for questions come on over to the Facebook tag :P we need all the flags you got! – Lix Feb 7 '12 at 21:10

I would say "no". See here for more stats graphs.

However, this one seems more worrisome:

  • 2
    Wonder how many of those get closed as dupes. – random Sep 6 '09 at 10:19
  • 1
    I haven't attempted to graph closed questions, that would be interesting too. – Greg Hewgill Sep 6 '09 at 10:23
  • Err, on that page, is the Vote Count Histogram horribly broken, or am I completely misunederstanding it? It shows one with 112k, and everything else on 0. – Macha Sep 6 '09 at 13:02
  • @Macha: For me it shows 112k having 0 votes, about 65k having 1 vote, about 30k having 2 votes etc. – Jon Skeet Sep 6 '09 at 15:23
  • @Greg: It might be worth doing "questions by week" to avoid seeing the "per week" spikes and troughs for weekends etc. – Jon Skeet Sep 6 '09 at 15:24
  • @Macha: Ah, I suspect you meant the first histogram, whereas I was looking at the second. Basically they're the same, but the first is massively compressed due to the wide range. – Jon Skeet Sep 6 '09 at 15:25

Even if they are, that's not necesarily a bad thing. I think it signifies that a lot of people who would have asked a question on SO are finding the answer on SO without having to ask the question. That's one of the benefits of a 200k+ question pool.

  • It depends on what you envision the future of SO to be. If the questions slow the community will die out and SO will be nothing more than a reference library. A really good reference library but shouldn't efforts be made to keep the community alive? – Spencer Ruport Sep 6 '09 at 18:55
  • 3
    That's the downside of being real aggressive about dupes, etc. Newbies come in, get shot down and leave. Why do you think that Usenet died? – duffbeer703 Sep 7 '09 at 0:23
  • Would you also predict that Wikipedia will die because people can't create duplicate articles? SO has been intentionally and explicitly designed with a wiki-like characteristic to it, and is not in its nature a conversation forum, like Usenet. Anyway, I blame AOL for killing Usenet. – Stu Thompson Sep 7 '09 at 14:39
  • 1
    Finding duplicates for specific questions can be hard. And tedious. Unless the title matches up with an existing question, i suspect a lot of these will always slip on by and just be answered over and over... – Shog9 Sep 8 '09 at 4:04

The only questions left seem to be edge cases of interest to only a handful of developers.

The long tail. Hmmm...that very well may be the case for c#, but I wouldn't know as they don't attract my attention.

I mention it because c# is overwhelmingly the most popular language to ask question for on SO. (Why? Well, because Jeff Atwood is a prominent blogger and c# coder. It makes sense.) From my MSO question last month, c# is clearly dominant on SO. Is there a desire, need or whatever to diversify the user base and questions? my thinking is that other languages are, relatively, not so well represented on SO. Future growth of questions will increasingly need to come from other languages. To do so, this means SO will have diversify its user base.

http://lanai.dietpizza.ch/images/Stack-Overflow-Top-Ten-Languages-Over-Time-small.png Huge 3.3MB image

Update as per Greg's request: Each Language as a Percent of Total Questions.

Very interesting, as it seems that c# is becoming even more dominant.

These graphs are really messy. I'm inclined to do them again but with only 3 to 5 series.

alt text http://lanai.dietpizza.ch/images/Stack-Overflow-Top-Ten-Languages-Over-Time-as-percentage-of-all-questions-small.png Huge 3.3MB image

  • Could you perhaps graph those languages as a proportion of the total questions, sort of like the Market Share graph at news.netcraft.com/archives/web_server_survey.html ? – Greg Hewgill Sep 8 '09 at 8:56
  • Good idea. Done. – Stu Thompson Sep 8 '09 at 11:27
  • what algorithm are you using for the trend lines? You should do one without the initial period, say from october onwards since the initial period is too chaotic. I would say from that graaph c# is, whilst still the top dog, not growing and even slightly decreased over the latter 6 months. Try a 10 day rolling average if you fancy blowing more bandwidth :) – ShuggyCoUk Sep 8 '09 at 11:53
  • algorithm? I right click the series in MS Excel for Macintosh, select "add trend line". Not really sure how it does that. Regardless, good point about excluding data prior to October. Maybe even later as October was unusually busy. Also, I think having weekly/monthly numbers would be more valuable. Sigh...so much to do... – Stu Thompson Sep 8 '09 at 12:01
  • Yeah, I was thinking aggregating by week might cut down on the noise a bit. Also, the "wall size" images are hard to view, but hey, I'm just one guy and could do it myself if I weren't such a slacker. :) – Greg Hewgill Sep 8 '09 at 19:42
  • Won't need the wall size if I get it down to weekly intervals. I'm a slacker too, so gimme a while to redo... – Stu Thompson Sep 8 '09 at 20:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .