An additional complication is that in the case of sms, a full link to reset passwords is not sent, but a short code. The client then asks the user to enter this code, however the client may now not know where to send the code. Should it send it to the identity server or the homeserver? Which sent out the code?
An additional complication is that in the case of SMS, a full link to reset passwords is not sent, but a short code. The client then asks the user to enter this code, however the client may now not know where to send the code. Should it send it to the identity server or the homeserver? Which sent out the code?
In order to combat this problem, the field `submit_url` should be added in the response from both the email and msisdn variants of the `/requestToken` Client-Server API, if and only if the homeserver has not sent out the entire link (for instance in the case of a short code through sms). If this field is omitted, the client knows that the link has been sent in its entirety and the verification will be handled out of band.
In order to combat this problem, the field `submit_url` should be added in the response from both the email and msisdn variants of the `/requestToken` Client-Server API, if and only if the homeserver has not sent out the entire link (for instance in the case of a short code through SMS). If this field is omitted, the client knows that the link has been sent in its entirety and the verification will be handled out of band.
If the client receives a response to `/requestToken` with `submit_url`, it should accept the token from user input, then make a request (either POST or GET, depending on whether it desires a machine- or human-readable response) to the content of `submit_url` with the `sid`, `client_secret` and user-entered token. This data should be submitted as query parameters for `GET` request, and a JSON body for a `POST`.
## Tradeoffs
If homeservers choose to not proxy the request, they will need to implement the
ability to send emails and/or sms messages. This is left as a detail for the
ability to send emails and/or SMS messages. This is left as a detail for the
homeserver implementation.
## Future Considerations
At some point we should look into removing the `id_server` field altogether and
removing any email/sms message sending from the identity server. This would
removing any email/SMS message sending from the identity server. This would
drastically reduce the amount of trust needed in the identity server and its
required ability. This is, however, a good first step.