Re: [GENERAL] pgAdmin4 needs information of v10 SCRAM authentication

On Mon, Apr 24, 2017 at 3:04 PM, Akshay Joshi
<akshay(dot)joshi(at)enterprisedb(dot)com> wrote:
> I have gone through this, but still facing issue to encrypt/change the database server password. In pgAdmin4 we have "Change Password" feature where user will enter the old and new password for the database server, now we will have to encrypt it (in Python) as per SCRAM standards and set it to the database.

By using SET password_encryption = 'scram-sha-256' and sending the raw
password you would be able to hash the password correctly. Or you
could just mimic scram_build_password() (routine in Postgres code to
generate that correctly).

> The example you have given in https://www.postgresql.org/message-id/76ac7e67-4e3a-f4df-e087-fbac90151907@iki.fi I have below questions:
> To encode the password you already have entry from pg_authid table which won't be possible for non superuser to access that table. How we can get that value from pg_authid table or do we have any other solution to this.

I don't understand this question, any user can update this field using
CREATE/ALTER ROLE, and the client has no need to know this value for
the exchange.

> For constructing the whole client-final-message, we need to calculate ClientSignature and ClientProof, which depend on the nonces, and is therefore different on every authentication exchange. How to calculate ClientSignature and ClientProof?
> Can you please guide me here, how can we achieve that in python.

You will need a C equivalent of what is proposed in fe-auth-scram.c in
the Postgres code to build the messages that are exchanged from the
server, see particularly calculate_client_proof() which describes step
by step the calculation of the client proof when building the last
message for the client. I didn't check in details, but the routines
are the same as in the message above. The format of the hashed
password has changed a bit since commit 68e61ee though.
--
Michael