1. I don't want others who have access to my client machine to be able to see a history of what I did online.
2. I don't want servers to be able to know anything about me except maybe my IP address.
It feels like tying these two together under one setting makes them both fragile. E.g. for scenario 2, I don't care whether a web page can use local storage as long as they don't have access to the data between sessions.
I'd much rather have two options - hide from the server and hide from your boss (or whoever). And maybe some UI to help me always hide from specific servers or delete all the artifacts from a specific session after the fact.
I try to wipe my drives and repartition every 30 to 60 days, with a full OS reinstall. The Virtual Machines I run with VirtualBox are even less persistent than the bare metal, often stateful for mere hours. I do not retain browser history, and I have only about 5 bookmarks, and trash my cookies and cache at least every day, multiple times usually.
But I want #2 for like 75% of the time. The other 25% of the time, that state almost never lasts 48 hours. When I go to bed, the current browser state dies forever. I usually have a hard time staying awake for 48 hours straight.
This means #2 will become 100% every 48 hours, with 48 hours being an extreme maximum lifetime for session data, and the true norm being 8 hours (9 to 5, each work day).
Considering that #1 & #2 are sure to intersect every 48 hours, dividing attention between them seems burdensome.
I don't want the New York Times to ID me, but I don't care if my wife knows I read it.
I don't want red tube to forget about my all access pass, but I don't want it in my browser history.
the times could switch to genuine browser/device fingerprinting and store that information server side. if they are careful enough so that false possibles are not possible (rather let a few slip through) then they could effectively control how much free access everyone gets.
the browser could even encrypt all their data by default. (but for non-incognito mode with a known key) it could then write the normal and incognito data in such a way that you can't even see that there is incognito data in there if you don't have the incognito key
The amount of encrypted data could be a side channel
even without that consideration, for things like disk storage, there is no reason[1] why incognito mode should have less access than normal mode. all websites should function as normal. the only difference is that in incognito mode everything is wiped once it is closed, and nothing is written to disk.
[1] ok, so the reason for the limitation is that the disk has to be emulated in memory because incognito mode must not write to the disk which could leave artifacts behind.
this makes me wonder if it is possible to detect a difference in timing for example when writing lots of data with an emulated disk vs a real one.