|
|
|
|
# MSC3122: Deprecate starting key verifications without requesting first
|
|
|
|
|
|
|
|
|
|
Currently, the [Key verification
|
|
|
|
|
framework](https://spec.matrix.org/unstable/client-server-api/#key-verification-framework)
|
|
|
|
|
allows a device to begin a verification via to-device messages by sending an
|
|
|
|
|
`m.key.verification.start` event without first sending or receiving an
|
|
|
|
|
`m.key.verification.request` message. (The last sentence of the 5th paragraph
|
|
|
|
|
of the Key verification framework in the unstable spec, as of the time of
|
|
|
|
|
writing.) However, doing so does not provide a good user experience, and
|
|
|
|
|
allowing this adds unnecessary complexity to implementations.
|
|
|
|
|
|
|
|
|
|
We propose to deprecate allowing this behaviour.
|
|
|
|
|
|
|
|
|
|
Note that verifications in DMs do not allow this behaviour. Currently, Element
|
|
|
|
|
Web is the only client known to do this.
|
|
|
|
|
|
|
|
|
|
## Proposal
|
|
|
|
|
|
|
|
|
|
The ability to begin a key verification by sending an
|
|
|
|
|
`m.key.verification.start` event as a to-device event without a prior
|
|
|
|
|
`m.key.verification.request` is deprecated. New clients should not begin
|
|
|
|
|
verifications in this way, but will still need to accept verifications begun in
|
|
|
|
|
this way, until it is removed from the spec.
|
|
|
|
|
|
|
|
|
|
## Potential issues
|
|
|
|
|
|
|
|
|
|
None.
|
|
|
|
|
|
|
|
|
|
## Alternatives
|
|
|
|
|
|
|
|
|
|
We could do nothing and leave it in the spec. But we should clean up cruft when
|
|
|
|
|
possible.
|
|
|
|
|
|
|
|
|
|
## Security considerations
|
|
|
|
|
|
|
|
|
|
None.
|
|
|
|
|
|
|
|
|
|
## Unstable prefix
|
|
|
|
|
|
|
|
|
|
No unstable prefix is required since we are simply deprecating behaviour.
|
