Hi,
Just a quick note that we recently started reporting more granularly about the user agents that visit Wikimedia projects (without any loss in privacy, of course). All the
nitty gritty details in phabricator. Summary here:
Before: We
were grouping too many small buckets into "Other", so we had a big "Other" bucket:
To illustrate, let's pretend we have a less popular browser with three different versions, being used at numbers just below our reporting threshold:
* Dan's Browser v1 - 0.9%
* Dan's Browser v2 - 0.7%
* Dan's Browser v3 - 0.8%
We used to just roll all of that data up into "Other" - 2.4%. What we're doing now is just including "Dan's Browser (Redacted) - 2.4%". So this has the effect of increasing the reported traffic share of browsers like Firefox or desktop OSes like Linux. The picture
now looks like:
We copied old data and made it accessible here (this is not being updated):
And we'll be updating the dashboards with new data, these remain at:
Do please take a look and let us know if you see anything weird. One last detail that I'll pull out of that long phab thread for those of you who read this far: we now have both "Other" and "Redacted". This might seem confusing at first, so I'll explain here and we can discuss:
- Other: this is truly now just representing user agent details that our User Agent parser has not been able to identify, so it really means "Other" not just a big bucket we're stashing data into
- Redacted: this is the label we're using when we're rolling up multiple pieces of data for privacy reasons. Essentially, user agents that get hit less than 10 times per minute are rolled up into "Redacted".