Re: [whatwg/fetch] Perform TAO check when reporting, using global's origin (PR #1422)

> I don't really understand the setup I'm afraid. Could you explain it?

We have to re-run the TAO checks when we know which global is gong to expose the RT entry. In actuality, we can this TAO check and drop the one we're currently running during a response (that's sort-of what implementations do today).

> * Does this require to keep all the intermediate responses in memory until "finalize and report timing" runs?

Only the bits of the response necessary for the TAO check (TAO header, origin, etc.). I can modify the TAO check algorithm to receive the minimal argument set instead. It should not have an observable difference.

> * Why does "finalize and report timing" now end up doing multiple TAO checks for everything, rather than just nested documents?

We can perform this final TAO check only when the global to report to is not the same as the original `client`. As above, it should not have an observable difference but perhaps it would more readable and optimized in advance.

> * Is the global in "finalize and report timing" the correct one?

Yes, it's the global that's going to expose the RT entry.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/whatwg/fetch/pull/1422#issuecomment-1105170189
You are receiving this because you are subscribed to this thread.

Message ID: <whatwg/fetch/pull/1422/c1105170189@github.com>