On 14.04.2017 23:32, hiro wrote: >> That should be done in a separate session manager process, IMHO. > > yes, that's what i mean. it just needs to demux the info from multiple > browser processes and then just write them to the file exactly like > opera does right now.
The interesting question is how the interface between the mux and the browser processes should look like. Just files in the same formats doesn't seem a good idea to me - complex and doesn't scale well. Maybe we could even move out the whole accessor logic into that dmux, (which then could also support various storage formats) - the browser just treats it like kind of an (special purpose) database server. Perhaps we could map it directly to fileystem semantics, the browsers can still run on their own (but now using their special fs layout), and the mux is just a tiny fileserver (eg. via 9P). The "big hog" (as I already called it) would then mount all the corresponding fs services (cookies, bookmarks, sessions, etc living in their own servers) before starting the individual browsers and acts as a supervisor. --mtx
