For me the verify by text worked, but you can click on a lot of very similar places and you get different results. For example if you click verify it forces interactive verification. If you click the sessions and then click a session you can verify individual sessions. Of course you can't non-interactively verify a users main key.

I'm also confused why each device is handled separately. I would rather I just share a key around (and ideally it rotates occasionally) and not share what and how many devices I have and what one I am using at the moment.

> but you can click on a lot of very similar places and you get different results.

Aha, didn't notice that, thx!

> I'm also confused why each device is handled separately.

Well, I can understand it more or less (I guess kind of similar to confirming in Whatsapp your multiple open sessions on different devices, to ensure that nobody is using something that you forgot/left behind?), but doing it this way is quite hardcore - on the other hand it could be that the whole thing is deeply embedded in the software's encryption principles/guidelines => it would probably still be ok, but it needs to be explained better, be more clearly accessible.

I guess that having a rotating key (with the software asking from time to time "do you want to accept key jf8k4d9k?") would probably be confusing for non-technical users and would probably generate uncertainty/anxiety/etc... ?

Losing the device is an interesting point. However I think due to the way that cross signing works they could use that device to sign new sessions anyways. They would also have access to key backups so I don't think that case is supported well right now.

For the rotating key it would be automatically signed by the previous key or master key so no user-visible change would be shown.

> However I think due to the way that cross signing works they could use that device to sign new sessions anyways.

So you don't think that if I cross-sign devices A and B, and then I would cross-sign devices B and C, if I would revoke B then C would automatically become invalid as well?

Kind of similar question about "key backups" (to which keys would device C have access to?).

(I honestly did not ever look into all these details - I was hoping that this would be covered by more clever people)

It would make sense but I'm not sure how it is implemented. I can also just imagine revoking old devices because I don't use them anymore or have reinstalled them. In that case I wouldn't want the things it signed to be revoked. (Really just saying don't trust this key for anything in the future, but past things are fine).

Maybe the best solution would be revocation after a date. So you can say "don't trust anything after {time-i-lost-the-device}" or "don't trust anything after {now}" and it does the right thing. However that could be complex to implement correctly in software. Lots of bookkeeping.