Mandate that the new device commits to a particular identity key

[hughns]

Originally from #4129, these are suggested changes to #4108:

• Mandate that the new device commits to a particular identity key (by setting the device ID to the public part) and proves ownership of it (via a combination of ECDH and HMAC-SHA256)

Co-authored-by: Damir Jelić poljar@termina.org.uk

[turt2live]

@hughns is there something which needs to be done to merge this?

[hughns]

Apologies, I meant to have this as draft as it is destined to be merged into MSC4108

[turt2live]

I'm not sure I understand the purpose of using a PR for this.

[hughns]

Having discussed this today with @dkasak and @poljar, the conclusion is that these changes should not be part of MSC4108.

Whilst "device ID = identity key" feels like a sensible direction to propose for the reasons stated, the implications of such a change are beyond the scope of the present (Sign in with QR) MSC.

We haven't been able to identify a benefit to the QR mechanism by having a proof of possession prior to sharing the secrets.

There may be an overall benefit in having the homeserver/OP enforce proof of possession during auth/API calls. This may mean that the check that the existing device does to ensure the new device is online before sharing secrets would be after the homeserver/OP has also approved the device.