Travis: I intend to be brief. I think this is amazing. It describes the typical lifecycle of a web page and it introduces 2 new states that a web page can go into and associated signals. The 2 new states are "stopped" (memory resident but not getting time on CPU) and "discarded" (forcefully terminated, gone) - when you see the stop signal you will be allowed to run a callback to unload some resources or prepare. After you've been stopped you may be terminated without further notice.

Dan: possible use of this to subvert user action..

Alex: the design point is that a lot of browsers are silently doing tab discarding - e.g. chrome on android, safari on IoS. They will keep snapshots (pixels) but the contetn is gone. Meanwhile a lot of content does want to live in the background - they will keep a lot of tabs open. What makes sense on mobile makes less sens on desktop - where those tabs might still be visible. That's one motivating use case. it's not designed to handle pop-up or pop-under detection on unload.

Travis: it seems the focus of the lifecycle is where the agent makes the decision to terminate. If the user closes the tab that's out of scope. Preventing ransomware pop-ups is high priority. I think it's a valid concern.

[ dbaron's machine spontaneously reboots and he disappears for a minute or two ]^{Citation needed}

Travis: will continue writing up feedback.

--