Turning off outdated skins/stats
This document aims to inform the discussion about switching off rarely-used and poorly maintained skins. Based on data from the Wikimedia Foundation database, our conclusion is that Chick, Standard, MySkin, Nostalgia, Simple and Modern are so rarely used as to only questionably justify support.
Methodology and limitations
The basic dataset we started with was a list of users grouped by skin choice, number of edits and activity levels. The groups were:
|Last active (months ago)||Edits|
The relevant data was retrieved from each Wikimedia project, and, excluding "non-project" projects (the checkuser wiki, the board wiki, so on) merged into a single dataset, which was then analysed using this script. (Note: this script no longer exists. Here are the latest numbers.)
One limitation of the data is that we could only retrieve users who had actively set a preference for their skin; users who had not done this displayed with a null value in the relevant table. Such users are not included in the numbers for Vector in the next section, which leads to an underrepresentation of Vector. Based on an approximate calculation, this may exclude up to 85 per cent of users.
While this sounds bad, it's actually not that big of a problem. Users who have not explicitly set their skin are automatically exposed to the default skin, Vector. In other words, while this lack of data does throw the numbers off, they throw it off in a direction we don't care about: the results of the analysis are biased towards, rather than away from, the skins we're considering removing.
Global skin usage
Based on the global dataset of user preferences, the use of skins for all accounts is:
|Skin||Total number of accounts||Percentage|
To summarise the impact of our changes, then; they will affect 151,578 accounts globally, or 1.05% of accounts. This doesn't tell the whole story on its own, however: most accounts overall are going to be inactive. If it turns out that only 1 per cent of accounts use to-be-deprecated skins, but they're 30 per cent of the active population, we have a problem.
In an attempt to gauge if this is an issue, the data was again analysed, but this time cut down to power users. For these purposes, we are looking at users who have:
- Been active within the last 6 months; and
- Have made over 1,000 edits, total, on their particular wiki.
Looking at that we get the following data, we see that any switch-off across all Wikimedia wikis would impact 397 power users (2.69%).
|Skin||Absolute number of power users||Percentage of power users|
- Modern (as is Cologne Blue) is supported by a particularly-enthusiastic volunteer developer, will probably continue to be accepted on Wikimedia projects.
- "Activity levels" refer to the last time a user logged in, or took any action logged in.
- Comparing the number of accounts on the English Wikipedia, which has the most account inflation, with the number of accounts that had actively set preferences.