Navigation Preload is an important Service Worker API allowing to offset the startup cost of Service Worker by issuing a parallel request for the main resource.
Typically, we've observed 300~400ms (1 RTT) wins on mobile with key partners (I'll try to get them to chime on this bug).
- Available in Chrome 59+
- Mozilla is supportive, tracking bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1290958)
- Edge is supportive 
- I believe that Apple was also supportive 
Navigation Preload is _one_ of the necessities for the Google Search team to ship a service worker on https://www.google.com/. Latency is a major concern, and Navigation Preload was the biggest win in several of the Chrome team’s experiments. For your consideration. Thanks.
The Google Search team have just documented why they need the Navigation Preload API: https://web.dev/google-search-sw/#service-worker-overhead.
Youenn, at the TPAC meeting today you asked that a feature request for service worker be filed. This bug already exists. Do you still need a new request to be filed?
Also, what information do you need here? At the meeting today I believe we discussed a number of large sites depending on this feature. This lists Google search, Linkedin, Facebook, and other MS properties like Outlook:
(In reply to Ben Kelly from comment #4)
> Youenn, at the TPAC meeting today you asked that a feature request for
> service worker be filed. This bug already exists. Do you still need a new
> request to be filed?
Thanks for following up.
This bug is good enough if all web developers are happy to put public information here.
> Also, what information do you need here? At the meeting today I believe we
> discussed a number of large sites depending on this feature. This lists
> Google search, Linkedin, Facebook, and other MS properties like Outlook:
Knowing that these websites are relying on navigation preload is useful for prioritization.
As discussed during the meeting, it would be really helpful to be able to measure perf benefits on some of those websites while working on further optimisations, including preload manager.
Seems like this was fixed via https://bugs.webkit.org/show_bug.cgi?id=233755.
(In reply to Thomas Steiner from comment #6)
> Seems like this was fixed via https://bugs.webkit.org/show_bug.cgi?id=233755.
Yes, it can be tested on the latest STP 138.
This has now shipped.