SageMath for Windows download counts via GitHub

[E. Madison Bray]

Something that has vexed me since first posting Windows releases to
GitHub is that I would have liked to have at least *some* download
statistics for it, and yet GitHub doesn't display that anywhere on the
website.

However, I was just noodling around in the GitHub API (trying to
automate releases) and I found that the data *is* there--or at least
some approximation thereof. It's probably difficult to give a precise
count since there are no doubt layers of CDNs involved, but the GitHub
releases API does have a "download count" for all release assets, and
just for the sake of interest this is what I found:

SageMath 8.0 (Windows installer 0.2) 16166
SageMath 8.1 (Windows installer 0.3) 25363
SageMath 8.2 (Windows installer 0.3) 7071
SageMath 8.3 (Windows installer 0.4) 4789
SageMath 8.3 (Windows installer 0.4.1) 3112
SageMath 8.4 (Windows installer 0.4.1) 4480
SageMath 8.5 (Windows installer 0.4.1) 3017
SageMath 8.6 (Windows installer 0.4.1) 3640
SageMath 8.6 (Windows installer 0.4.2) 5862
SageMath 8.7 (Windows installer 0.4.3) 2929
SageMath 8.8 (Windows installer 0.5.0) 101
SageMath 8.8 (Windows installer 0.5.1) 4139


Apparently there was a flurry of interest when Sage for Windows was
first released, and then it tailed off a bit. However, it's worth
noting that previous releases were not also on the sagemath.org
mirrors, whereas they were added to the mirrors later (probably, just
looking at the numbers, around 8.2 or 8.3).

So those number just reflect downloads through GitHub, but at least
it's something.

[rjf]

Just curious -- if sourceforge says there are 65,766 downloads of Maxima 5.42.2 (say)

for Windows, does that mean there are really another 4139 downloads of (some version of)

Maxima that are not counted because they were downloaded from SageMath?

I assume Maxima downloads for linux are undercounted because people

get them from some indirect distribution.

Maybe this counting of downloads could be done in some better way.

What comes to mind is some feedback from installations, e.g.

running the installation sends email somewhere to register it.

But that would be kind of intrusive, and also assumes internet

connected + email.   

Of course it is also difficult to judge how many copies are

running because some people will continue to run an old

version, others will repeatedly load the system to get the

latest fixes ...

Good luck on trying to count "users" instead of copies.

RJF

[kcrisman]

Good luck on trying to count "users" instead of copies.

RJF

True that.  Though I think it is interesting that in practice we get many contacts from people who do *not* download new versions very often, so there is at least some vague connection between downloads and users (as opposed to people simply upgrading with every release).   How often they use it is another story - which is also the case for proprietary math software, I can think of many places with site licenses for an M* that use it in one or two courses, and only if certain professors teach it.