Re: SCRAM methods

[Simon Josefsson]

>On 14/01/2020 22:19, Simon Josefsson wrote:
>> Please try version 1.9.1 and tell me if it does what you
>> want!  There are new properties for ServerKey/StoredKey now.
>
>Yes, that works nicely for the server side.
>Client side in the next patch release?

Currently I prefer to not implement ClientKey support -- there isn't
any security advantage with it as far as I can see.  Stealing
SaltedPassword or ClientKey/StoredKey both make client and server
impersonation possible.

I believe clients should store cleartext-password or the
SaltedPassword.  Storing PBKDF2 values in clients is not that uncommon,
so there may be infrastructure in place for it in some environments --
whereas ClientKey/StoredKey are entirely SCRAM-specific.

I could be convinced to change my mind, but currently I value lesser
code complexity both in gsasl and in application as more important than
any perceived advantage with ClientKey/StoredKey over SaltedPassword. 
I just don't see any significant advantage with ClientKey/StoredKey
over SaltedPassword.  It saves two HMAC operations is all I can see,
but those are cheap compared to code maintenance.

Thoughts?

/Simon

signature.asc
Description: This is a digitally signed message part