Talk:Reading/Web/Projects/Performance/Lazy loading references

Usage

Why do images and references at the same time, rather than just images? How do you test whether this change increases the time needed to reach a reference, counting from the moment one clicks the superscript or reaches the section? How to make sure you don't break JavaScript which loads the reference "inline"? How to check the long-term effects, e.g. whether in the long term you make the users less likely to use the references? What can we do to optimise for maximum usage of the references, rather than for minimum use of them? --Nemo 23:14, 9 August 2016 (UTC)

While we are building these two solutions at the same time, we are rolling them out separately. Right now we have been focusing on deploying and measuring lazy loading images. With respect to lazy loading references we have built a solution that works without JavaScript, you are right that there are other things to think about here. It's worth pointing out that Tilman's analysis showed low usage of the reference section. Usage of actual references (via the popup dialog) remains unchanged and is probably more popular (although we'd need to re-run EventLogging to guarantee that's the case). Right now we are exploring the performance gains and bytes changed of deferring their loading before investing more energy in this endeavor. Jdlrobson (talk) 15:28, 10 August 2016 (UTC)

Of course everything we want to encourage is done rarely, otherwise there would be no need to encourage it. Wikipedia exists to teach people to check their sources. Nemo 15:38, 10 August 2016 (UTC)