Added navigation-timing/nav2_test_redirect_chain_xserver_final_original_origin

[nicjansma]

Adds a test that:

• starts from a page on web-platform.test/...
• iframe added for www.web-platform.test/... (cross origin)
• which redirects to www.web-platform.test/... (cross origin)
• which redirects to web-platform.test/... (same as original origin)

In this case, because of the cross-origin redirects, redirectCount, redirectStart and redirectEnd should all be 0.

[nicjansma]

CC @yoavweiss @npm1

[npm1]

Looks great!

[sefeng211]

Question: redirect-opt-in.py sets TAO header for the cross-origin redirects, shouldn't it allow timings to be returned?

[npm1]

Looking at the deprecated interface, it seems to imply that the redirects should not be set for cross origin redirects (regardless of TAO). But I don't know where the create the navigation timing entry is called, do you know @yoavweiss or @noamr ?

[noamr]

Looking at the deprecated interface, it seems to imply that the redirects should not be set for cross origin redirects (regardless of TAO). But I don't know where the create the navigation timing entry is called, do you know @yoavweiss or @noamr ?

It is called from the HTML spec.

For navigation timing, redirectCount is set to zero if there are any cross-origin no-cors redirects here.

The resource timing properties (redirectStart and redirectEnd) are handled in FETCH with the same TAO rules as any resource.

[npm1]

Ok thanks, does that answer your question @sefeng211 ? I also filed w3c/navigation-timing#157 because it is really hard to find answers in this spec now.

[sefeng211]

So er....this test was written based on the deprecated interface, but it should be interpreted based on the current Fetch spec right...Am I misunderstanding something?

If so, should I read it as, it returns 0 because the redirectCount is 0? I am still not sure why we should return 0 for redirectStart in this case.

Sorry! I am having a hard time to understand the spec..

[npm1]

Ok, so I think the spec is currently following what Resource Timing is doing, which is to set it depending on TAO checks. In particular this test and the spec seem at odds. I filed w3c/navigation-timing#158 to find out the answer, apologies for the mess in this spec!

[noamr]

Ok, so I think the spec is currently following what Resource Timing is doing, which is to set it depending on TAO checks. In particular this test and the spec seem at odds. I filed w3c/navigation-timing#158 to find out the answer, apologies for the mess in this spec!

Right. something is actually missing in the spec. I'm on it, thanks for capturing it.