Re: Fwd: "POP3 SASL Authentication Mechanism" submitted for publication

Frank Ellermann wrote:

>Alexey Melnikov wrote:
>  
>
>>>>The multi-line response client to server confuse me.  
>>>>Which SASL mechanism needs this ?
>>>>        
>>>>
>>>There's no multi-line response (unless I'm missing something).
>>>      
>>>
>>Indeed.
>>There are mechanisms with multiple challenges/responses.
>>    
>>
>Let's see, I hope I got it now.  What really happens is this:
>
>C: AUTH mech initial-response-if-allowed-for-mech
>S: + challenge
>C: response
>S: + challenge
>C: response
>S: +OK your're logged in, maibox locked, have fun
>  
>
Correct.

>However the ABNF put's the complete part of the client into one
><auth-command> = "AUTH" mech [SP ir] *( CRLF [base64]) CRLF
>
>My confusion was that I thought the client sends this complete
>multi-line <auth-command> at once, without intervening server
>challenges.
>  
>
Right.

>Maybe it's only me, then forget it.  Otherwise the ABNF has a
><continue-req> for the "+" SP [base64] CRLF from the server,
>it could similarl also define a <continue-response>:
>
>auth-command     = "AUTH" mech [initial-response] CRLF *(response)
>initial-response = SP (base64 / "=")   ; a single "=" if empty
>response         = [base64] CRLF       ; after server challenge
>  
>
I don't object to something like this.

>  [Abhijit Menon-Sen wrote:]
>  
>
>>>there's no very good way to express this in the ABNF
>>>      
>>>
>Yes, but maybe using an explicit <response> with a comment helps.
>
>For Hector's multi-line observation I'm not sure what that was,
>an implementor confused like me, or some kind of pipelining.
>  
>