[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: HEADS UP: API error code changes

At 01:42 AM 12/11/2003, Hallvard B Furuseth wrote:
>Kurt D. Zeilenga writes:
>> Since protocol codes must be non-negative, I've changed all API error
>> codes to be negative values.
>Negative codes make me a bit nervous.  Why not use codes >= 16384, since
>they are for private use according to rfc3383?

Because nothing prevents a server from returning those codes.
Even if we register the codes, we'll get leakage (just as we
do today with the registered RFC 1823 codes).  By going to
negative numbers, if there is leakage, it will be clear that
the server is violating the protocol.  As it stands now,
much of the current leakage could be viewed as valid per
the protocol.

While I fully expect there to some minor, one-time breakage
(most any change would cause some breakage), I think the benefit of
solving the ambiguity (is unknown code X an API error or a
protocol error)/leakage problems (got API error from protocol)
fixes behavior which is causing on-going problems.