Authorship & Access List

[edemaine]

Redesign for access control (with significant input from dev team including Yevhenii, Alice, Ankur, Dylan, Marty):

• authors mapping remains the ground truth of edits as reflected in history of edits, so recomputeAuthors will still do the right thing: map each user that is among the updators of some edit to the last updated date among these edits. (This makes backward compatibility easier, and makes it easy to remove and then re-add an author (e.g. accidental removal) without having to dive back into history.)
• Every message has a coauthors list/array (like current authors, but no longer necessarily corresponding to diff sequence).
  • Being on this list implies full message access (like authors used to), regardless of flags.
  • Existing authors and @mentions will be converted into this during bootstrap process.
    • ...and bootstrap updates to coauthors tracking throughout history
  • New authors are also added to coauthors (via $addToSet).
  • Changes to coauthors should be tracked in diffs, and reported (as adds/removes) in notifications, just like @mentions would have been. This will encourage good behavior.
    • Perhaps special notification if you've been added or removed from authorship on a message by other people (i.e. by an edit that wasn't yours).
  • (Multikey) MongoDB index (Efficient access detection #149). Actually, unclear how an index would help here, except for author view, but is that worth optimizing? Likely being replaced by a search view.
  • authors no longer gives permission (replaced by coauthors)
  • @mentions no longer give permission!
    • This fixes the overloading of @mentions for both referring to people (and thus want notifications, like At mentions should do something to email notifications #384) and for permissions/authorship (for tracking who worked on something, pretty much the opposite of "want notifications")
  • @mentions no longer important in statistics view (at least with respect to zeroes -- but still might be interesting to draw)
  • @mentions no longer seem important for bottom message summary (next to coauthor count) -- remove? Or perhaps replace with links to @mentions to yourself?
  • Update MessageAuthor to show coauthors that haven't made explicit edits in history (Condense editor list #393)
• In editing mode, at bottom, have boxed author list, and add a "+ Author" button or box
• Authors have remove button, similar to tags. Useful for fixing mistakes (e.g. student edits an accidentally non-protected post and gains authorship, but see Protected/locked/restricted flag for messages #507).
  • Should have confirmation dialog / be inside the user hover (via click to popup instead of hover): UserLink popups #510
  • Superuser can remove any authorship
  • Allow people to remove their own authorship even if they edited (e.g. when fixing a typo, though don't necessary encourage this behavior as it's kinda micro-managing authorship).
  • Allow people to remove other coauthors that aren't authors (i.e. didn't literally edit the message). This lets scribes who accidentally added the wrong coauthor fix their mistake. Notifications about changes should encourage the right behavior.
  • This is where coauthors list can deviate from authors list, and why we don't want to give authors permission anymore (replacing with coauthors).
• Maybe an Action / Coauthor button so spectators can easily add themselves to the coauthor list without entering editing mode, entering their name, and exiting editing mode?
• Should there be an Action / Uncoauthor? (with similar confirmation dialog) No, terrible name, and seems like a rare action; if you need to fix a mistake, fine to go into Edit mode and remove your name there.
• Every message has an access list/array.
  • Being on this list implies message access unless message is unpublished or deleted. This is effectively useless for public messages, but useful for private messages.
  • Should indicate access isn't effective for unpublished or deleted messages in access list UI.
  • Should be visible in message view, not just editing view.
  • (Multikey) MongoDB index (Efficient access detection #149). Actually, unclear how an index would help here...
  • New line in statistics view?
• In editing mode of private message (or message with nonempty access list?), add another line of "The following people can also access the message:" [or will be able to once published/undeleted] with differently colored boxes or something.
• Private messages get Reply All button which makes a reply with access list = parent's access list union parent's author list (but not synchronized). This makes it easy to have back-and-forth private discussions. Replaces Inherit permissions from parent #475.
• ?? Reply None (or Reply to...), or Reply to Staff, or Reply to Authors with less permissions
• Probably should be able to post private reply to a private message you're given access to, if you can see the private message (independent of thread permissions) -- or is this already the case? Yes, we already inherit from the parent; only forced by the root when the root says "private only".
• @mentions should give suggestions for adding as author (if message is public or private) or access (if message is private). Usually you want a mentioned person to read the message you're mentioning them in, so this will help fix that error, but sometimes you don't, so it will also not force it.
• Allow superusers to add non-group-members / arbitrary username strings in UserInput? This was possible with @mentions, but now it isn't possible. Superuser case is useful for prewriting a set of problems before inviting everyone to a group.
  • Should we allow anyone to do so too? Relevant in a write-only group (WriteOnly demo explicitly mentions you can add people via @mention), and I guess for users with global read access (but that seems less important).
• Add hover help to "Coauthors" and "Additional access"
• Maybe History view should show "shadow" edits: edits that are logged in authors but not coauthors
• Perhaps disable removal of creator's authorship, except by superuser
• Removing yourself as an author should also stop you from editing the message
• Shouldn't show access list when message isn't private. (It's fine that the access list gets remembered in case it gets made private again, but it's weird to see "by ...; access to ..." in a public message.)
• Unpublished private messages should maybe have "eventual access to" even when not in editing mode, so you know what will happen after "Action / Publish".

Replaces #177, #475, #149, #393