Hi blink-network-dev team and platform-architecture-dev team,
Context
We are planning an infra migration for fetch keepalive requests. The current implementation suffers from delaying the shutdown of a renderer process for an extra 30 seconds. The plan is to move request loading into the browser process when the renderer is unloaded.
Design Doc
Questions
* From yhirano's notes, there might still be some security checks that are not yet moved to //network/service. Maybe I need to servicificate them first. Could anyone help point them out?
* It is unclear to me whether the fetch response could just be dropped if it's loaded in the browser process.
* I understand that the fetch API is a very critical feature, but I am not very familiar with the fetch/loader module in the blink, so there might be something missing from the doc. Please help correct my proposal.
Thanks!
Ming