Seems to still be an issue as we just had a user showing up on IRC reporting it.
Trac: Summary: Regression: Torbutton 1.6.6.0 will not save Exceptions in the Firefox cookie manager to Tor Browser will not save Exceptions in the Firefox cookie manager
But this is by design. Permissions (all site exceptions are kept in permissions.sqlite) are disabled from writing to disk, and are memory only
permissions.memory_only = true
I feel like I'm missing the point, since this ticket has been open for 6 years
Originally, the ticket has meant something slightly different as Torbutton back in the day had the option to save cookies across sessions if the user chose that option.
However, even if that's not the case today anymore the current situation is still a bug. We need to decide where the bug is, though (I am not sure what the expected behavior in vanilla PBM is but I assume cookies can retained across sessions. Maybe should follow that model? Or maybe we should adapt the UI if permissions.memory_only is set to true making it clear there is no across-sessions-thing. Or maybe the bug is something else). However, offering the option to save exceptions across restarts and not following along to the surprise of users is clearly sub-optimal.
We need to decide where the bug is, though (I am not sure what the expected behavior in vanilla PBM is but I assume cookies can retained across sessions.
No. Cookies (and without double checking, sessionStorage, localStorage) in PBMode are memory only. You can test by checking the cookies.sqlite file in normal mode vs PB mode
Moot anyway, since this is about permissions, not the actual persistent data :)
The distinction here is persistent "web data" vs "user data/settings" and they are different threat models (browsing the web vs having your OS compromised): e.g. PBMode allows bookmarks, passwords, site exceptions etc to be retained, but not history (AFAIK: there have been changes to PBMode in this regard since 68, but I'd have to dig them up). We don't stop people creating bookmarks for usability reasons, so why should we stop other "user" data.
I honestly think this should be distinguished (web vs user) and relevant pref(s) flipped - maybe in the slider (but permissions.memory_only at least requires a restart = too messy)
At the very least, be consistent about disk writes: because it seems like a mixed message here.