Set request's history-navigation flag for history traversal

[yutakahirano]

See w3c/ServiceWorker#1167 for the discussion that led to this change and whatwg/fetch#718 for the infrastructure in Fetch this builds upon.

Tests: web-platform-tests/wpt#10909

---

/acknowledgements.html ( diff )
/browsing-the-web.html ( diff )
/history.html ( diff )
/infrastructure.html ( diff )

[domenic]

Editorially this looks fine, but I have a hard time telling if the step is inserted at the right place to cause the intended effect. For example, I'm unsure whether we want this flag set for all navigations to new HTML documents (which invoke "update the session history with the new page" and thus this algorithm), and I'm unsure whether we want to exclude setting it for bfcache navigations (i.e. cases where step 1 of this algorithm is not true).

@annevk, I thought you were more involved with this; can you help confirm?

[annevk]

Well, I'm still skeptical about the whole navigate algorithm being correct, but you all wanted to move ahead with these things so changes to HTML I'm trying not to say too much about.

[annevk]

(Those do sound like problems though.)

[yutakahirano]

Thank you!

For example, I'm unsure whether we want this flag set for all navigations to new HTML documents (which invoke "update the session history with the new page" and thus this algorithm),

You're right. What I was trying to set the flag was requests via traverse the history by a delta, so I fixed the change.

and I'm unsure whether we want to exclude setting it for bfcache navigations (i.e. cases where step 1 of this algorithm is not true).

If entry holds a Document, Navigate algorithm is not called and service workers cannot see the request, right?

[yutakahirano]

@domenic , what do you think about the latest change?

[domenic]

Sorry for the delay; I had TC39 this week. The latest version looks good to me. I made further editorial tweaks around how the different flags to that algorithm are used; we don't have a fully consistent style across the spec, but this one seems a bit better.