Goal:
Quickly gather information about engagement with coordinates and maps on mobile
What:
- Enable the gadget for everyone (anonymous and users) for a 2 to 4 week run.
- Gather data on engagements with the coordinates and the map button on mobile web (and overall tile usage)
- Announce on en.wp village pumps ?
Gaps in approach:
- No full design cycle (might not be the most enticing way to have users interact with this information)
- JS based
- No coverage of mobile apps
- Probably hard to compare to desktop numbers, due to multitude ways that we offer desktop users this information
Benefits:
- low impact on desktop users
- low investment by both WMF and volunteers
- big enough pool of users to get usable data from
Previously raised concerns
So to reiterate:
- Payload size concerns
- Alleviated by avoiding oojs ui widgets
- Only affected location pages, which should be 4% of our pageviews (according to JKatz)
- Concerns regarding Sierra Leone tiles
- Could be a Kartotherian issue
- Could be a connection problem
- Could be an overload due to increased usage ?
- I do not see a major increase in resource usage in grafana, though it does seem there was a half to 1 million served tiles 'bump' when the gadget was active.
- Regardless, this seems likely to be serverside, if anything, so not directly related to the gadget
- Amire reported issues with Firefox
- These issues cannot be reproduced
- No logging is available from the problems.
- Gadgets don't work for anymous mobile users
- We can avoid this by loading from MediaWiki:Mobile.js instead
- The default gadget throws a non-problematic console error
- We can avoid this by loading from MediaWiki:Mobile.js instead
- This issue is fixed now https://gerrit.wikimedia.org/r/#/c/376036/
- Is ops aware of increased tile server usage and could it be a problem
- Ops is aware now
- Everyone seems to agree that the increased usage should not be a problem
- The addition of the map logo causes a repaint and a reflow
- The reflow has been dealt with
- The repaint can only be solved by moving the code completely into the extension
- For a limited time, it might be reasonable
- Might even increase engagement, by the button being more noticeable.
- What's the impact for lower end devices ?
- I would say it's the same as for people visiting wikivoyage pages with maps.
- Do we have any defences to these kinds of problems to begin with ? Can we identify a performance group and exclude them from certain functionality ?