On github maybe...

Github isn't the only source of code available on the web.

MIT is loved by business, that can take the code from others, packaged it with a new shinny face, sell it and don't return anything back.

But one discovers who they are when their customers have problems.

On my local machine (since I have texlive installed and each texlive package is a separate package in openSUSE ignore the fact that `LPPL` is the most popular). GPL-2.0+ is ~2x more popular than MIT (and all of the *GPLs together are ~4x more popular) in my admittedly small sample size of packages.

    % rpm -qai | awk '$1 == "License" { print $3 }' | sort | uniq -c | sort -nr | head -n10
        1248 LPPL-1.0
         558 GPL-2.0+
         301 MIT
         293 LGPL-2.1+
         221 BSD-3-Clause
         138 GPL-2.0
         131 GPL-3.0+
         115 SUSE-Public-Domain
          79 Artistic-1.0
          41 Apache-2.0

And actually I haven't counted dual-licensing. Here's an even better script:

    % rpm -qai | awk '$1 == "License" { print $3 }' | grep GPL | wc -l
    1316
    % rpm -qai | awk '$1 == "License" { print $3 }' | grep MIT | wc -l
    302

The link provides the API on how to poll github repositories for license data, but is a far cry from being a reliable method in measuring usage.

Lets suggest following criteria for a survey:

1#, no duplicates. A survey with duplicate has no meaning. Github data in the blog post do not exclude dubplicates.

2#, no empty projects. A project with no data in it is not actually a project. The Github data do not exclude such projects.

3#, only project with code in them. A lot of github projects is text collaboration, like party planning, discussion around the latests online drama, and so on. Github data do not exclude those.

4#, only project with running code that is useful for at least 1 person. Random data, incomplete or abandon attempts is not actually "a software project". Github data do not attempt to run projects, and very much do not limit its data storage to only code that works.

To my knowledge, only one license survey exist that passes those criteria, and that one was based on the curated repository of Debian. The study showed GPL to be heavily favored (80%), but the study concluded that more such studies (and discussion on research methodology) is highly needed.