| Commit-Queue | +1 |
It's been a while, but would like to get back to this cl, as it makes progress in getting the commit time to the browser. PTAL! :-)
base::TimeTicks commit_time;Johannes HenkelDo you think we could include both the URL update time and the first paint time (not presentation), and then commit_time would be the max() of these two?
I think this gives us two featues:
1. We could compute how often we get paint before url vs url before paint.
2. We could experiment with slicing on URL change immediately, even before first paint.Either way, we need to wait for FCP/LCP to arrive (i.e. presentation time) in the subsequent message before this soft-nav become "usable data", so the effect of measuring URL time without paint is equivalent to measuring before presentation time (for soft-LCP).
The effect on CLS/INP is unclear. I think there are cases where it would be better and some where it could be worse.
Sorry, took me a while to chew on this. But, now that we've settled on this particular definition of 'commit time' for soft navs, I think it's best to have commit_time as a field.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |