Correct TTFB breakdown

[tunetheweb]

Makes the following changes:

• Redirect time cannot be accurate measured so renames it to "Redirect and waiting duration".
• "Redirect and waiting duration" should be measures from workerStart OR fetchStart to handle non-SW pages.
• All times should be clamped to 0 after subtracting activationStart to accurately measure TTFB from activation for prerendered pages (i.e. from when user clicks rather than from when browser starts prerendering).
• This is often a gap from connectEnd to requestStarts which is currently being attributed to SSL time. Make the following changes:
  • Change SSL duration to measure from secureConnectionStart to connectEnd to more accurately measure this.
  • Measure request from connectEnd rather than requestStart so the gap is included in request time instead.
  • This is necessary as service worker requests do not allow measurement of TCP or SSL time so they should be 0 so better to attribute to request time than SSL time.
• Changes all "times" to "durations" to make clear they are spans of time and not actual times.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
webperf-snippets	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Apr 12, 2024 1:21pm

[vercel]

@tunetheweb is attempting to deploy a commit to the Joan Len's projects Team on Vercel.

A member of the Team first needs to authorize it.

[nucliweb]

Thank you @tunetheweb 🙌