Last modified: 2014-04-21 18:51:53 UTC
If possible, it would be good to know the number of users who are using a 2G/GPRS connection as their primary network connection. One way would be to find non-mobile User-Agents being used on netblocks reserved for Mobile traffic. This sort of use is more common in less developed areas that have cell coverage, but few to no cables laid to the home. From an email to Erik Zachte (after which he suggested this bug): We would look at *only* non-cellular UAs and find out how many of those were on clearly celluar subnets like Orange TUNISIE (www.orange.tn) which owns '41.224.64.0 - 41.224.127.255'. This would mean matching UAs to IPs. Alternatively, it would be useful if we could find out how much bandwidth a user has, but I'm not sure how to do that without being really invasive.
Prioritization and scheduling of this bug is tracked on Mingle card https://mingle.corp.wikimedia.org/projects/analytics/cards/1257
This is not Wikistats related. I have no tools for, nor experience with, determining particulars of any subnet. Does any other team member have clues on this?
I changed it to Wikistats because Mark mentioned that he had emailed you about this, it could be a feature request. It's definitely not User Metrics either. @Mark: to which component does this bug belong?
(In reply to comment #3) > @Mark: to which component does this bug belong? I'm not sure where you should put this, so let me explain why this information would be useful and you can direct me to the right place. Most of our developers have high-speed internet connections. However some users use slower connections. Without a way to track this information, we remain oblivious to their concerns until a saavy enough user complains. This specific bug requests information about people using their mobile connection as their main point of access to the internet. This is important because mobile developers may tend to dismiss this use case as insignificant. Maybe they would be right. But until we have data, we have no way to measure the significance. The larger issue, though, is that we should have data on users and their connection speeds. Until we have this, we have no idea what percentage of users find the "heavier" interfaces that we're providing too unweildy to use. Also see: Bug 55842 - UI tests should check performance on slow connections