[ietf-nntp] Re: AUTHINFO/SASL responses

[Russ Allbery]

Clive D W Feather <clive at demon.net> writes:
> Russ Allbery said:

>> Because I really don't want to invent a new response when we don't need
>> to.  If we can't use the standard response code for a syntax error,
>> let's use the existing error message for bad auth protocol exchange.

> I don't know about "need to", but it strikes me that it would be helpful to
> the client to distinguish:

> 1   AUTHINFO JUNK
> 2   AUTHINFO SASL INVALID
> 3   AUTHINFO SASL VALID JunkBase64
> 4   AUTHINFO SASL VALID Base64notNeeded=

> 1 and 3 are syntax errors, but of different kinds.

Actually, 1 and 2 are identical cases as far as I'm concerned; they're
both unsupported subcommands (501).

3 is a syntax error, yes.  Unfortunately, we use the same error code for
syntax errors and unsupported subcommands.

4 should definitely be a different code.

> 2 and 4 are semantic errors of different kinds. If you're trying to
> debug a problem and you get the same error code if the far end doesn't
> do SASL, or does it but doesn't have the FOO-SPECIAL module, or has the
> module but you've mistyped the base68 encoding, it does not make things
> easy.

Well, this is a general question about debugging commands with
subcommands, isn't it?  If you run a hypothetical:

    LIST CHARSET 1- Subject

and you get back 501, did you forget the order of the article number and
header arguments, or did the server not support LIST CHARSET?  Well, LIST
EXTENSIONS will tell you.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>