Reliability of extension stats

@jcfr @Sam_Horvath

How reliable old download statistics? I just ran the extension download statistics, for all extensions, and why I sort them grand total, I am a little surprised with the some of the top results, such as SlicerXNAT, which I believe is not even maintained.

Most of these packages very high download stats for version 4.10.2 (in the orders of 50-80K), which kind of looked a bit superfluous to me, particularly given their recent usage has been quite low.

Downloads for both the Slicer built-in extension manager or the corresponding web page loaded in a regular browser are accounted for.

We are not discriminating between both and we are not performing filtering like it has been implemented for the Slicer download packages.

Possible ways to improve this and tease apart the “source” of the download would be to either customize the user agent associated with Slicer built-in browser and/or append a new parameter.

Oh, I see if the source repo like GitHub - MokaCreativeLLC/XNATSlicer: XNAT-Slicer Integration is accessed from a web browser, that still counts as a download?

I don’t think it is necessary to distinguish download count obtained from the web browser or from the extensions manager. When someone downloads a package via the web browser, the intent is most likely to install it in Slicer.

This does not count as a download. Only extension package downloads are counted.

The numbers look good to me, I don’t see any reason to doubt their accuracy.

Slicer has a long history, so if a more recently added extension needs to be compared with older ones then you may consider limiting your analysis range to releases from the past few years.

The main limitation of the download count that it does not reflect usage. For example, a completely useless extension may get larger download count if its name suggests that it could be useful.

We would need to collect usage statistics to more accurately measure impact or usefulness - see Should we start collecting software usage data?

@muratmaga What is meant by this is that both from within the Extensions Manager web browser inside the 3D Slicer application or from https://extensions.slicer.org/catalog/All/31938/win which can be accessed by the instructions at https://slicer.readthedocs.io/en/latest/user_guide/extensions_manager.html#download-extension-packages.

@muratmaga you just spent a week with a bunch of users at the workshop, did you happen to get a sense of how they would feel about the question of sharing anonymous usage statistics? If not, could we reach out to them and ask?

I think most people would be okay with something along the lines of the RenderDoc example (Should we start collecting software usage data? - #24 by jcfr).

We did not discuss this (in hindsight would have been a good topic, but then we already have a good sense what modules Slicermorph users are interacting with).

I think most people would be ok with anonymous usage stats provided:

  1. It is not intrusive and enabling the telemetry does not slow the work.
  2. Clear indication of they would benefit from this.