Skip to Content.
Sympa Menu

commons-research - [Commons-research] Quantifying publicly indexable software using Google Code Search

commons-research AT lists.ibiblio.org

Subject: Commons-research mailing list

List archive

Chronological Thread  
  • From: "Ben Bildstein" <ben.bildstein AT student.unsw.edu.au>
  • To: commons-research <commons-research AT lists.ibiblio.org>
  • Subject: [Commons-research] Quantifying publicly indexable software using Google Code Search
  • Date: Fri, 2 May 2008 14:46:23 +1000

Hi, once against I'm list/blog cross-posting, but sue me: there's not been a lot of action on this list recently anyway :)

Okay, so here's the gist. I gathered a little data using Google Code Search (Advanced). I wanted to know how the various open and free licences compared. Here's a spreadsheet I created after querying a bit: http://spreadsheets.google.com/pub?key=pc2Aa53AzQmzqDINCGvhuZw.

Now, the results of this.
  1. You can get an approximation of the whole index by doing a regular _expression_ search. Approximate total (non-unique) source code files indexed: 31,000,000.
  2. You can't get access to more than about 200 results. It might tell you there are more, but only the first 100-200 are shown.
  3. Google's estimates are really, really bad, and depend on the order of magnitude of the results. For example, results for the Apache licence changed from 1.55 million to 156 (yes, they shrank by a factor of 10,000) in 5 days. Oh, and today it's back up to 815,000, though you can actually only get the first 166 of them.
  4. The (logarithmic) graph shows a severe discontinuity, and that is precisely the point where the graph changes between approximate "of about" results and specific results. This suggests to me that either the left 6 points should really be down lower by a factor of 10,000 (3 on the logarithmic scale), or shifted by a fixed 140,000 or so, or the 19 or so points on the right should be much higher.
The data would have been interesting, but the results are so variable that what we have instead is another interesting statement about search engines' ability to quantify search results.

Ben.


e: ben.bildstein AT student.unsw.edu.au
b: http://cyberlawcentre.org/hoc
m: 0408 134466



  • [Commons-research] Quantifying publicly indexable software using Google Code Search, Ben Bildstein, 05/02/2008

Archive powered by MHonArc 2.6.24.

Top of Page