Proposal: Restricting the font-display

[yshrdbrn]

I opened this issue here to discuss different aspects of the feature proposed in w3c/webappsec-permissions-policy#269 including the usefulness of it. I am currently working to see if I can get statistical results of how much turning off the feature will affect the layout changes. I will post the results as soon as I have them.

P.S. any suggestion for a better feature name is much appreciated!

[yshrdbrn]

Test Results

Out of about 200 websites that use webfonts, There is about 9.5% improvement in layout stability.
By using the Layout Stability Metric, Each website was given a Jank score.
The improvement is obtained by this formula:

PNE = Jank score when policy not enforced
PE = Jankscore when policy enforced
Percentage improvement = (Σ(PNE) - Σ(PE)) / Σ(PE) *100

Here are the result charts:

Most improvements happened when the Jank score was in the range (0.1, 0.5). We see 29% improvement in that range.

Data Collection:

Each website was loaded 15 times on a simulated 3G connection with the cache disabled. The median of the samples was collected as the score of each individual website. The test was run locally on one machine (loading one website at a time).

[Malvoz]

I just stumbled upon WICG/interventions#58 - I think a conclusion for that intervention may affect the usefulness of this proposal.