RE: [NSIS] Error codes

["McDonald, Andrew" <andrew.mcdonald at roke.co.uk>]

Hi,

Geib, Ruediger wrote:
> I appreciate that the error coding is done as learned from
> other state-of-art protocols and support your proposal.

OK.

>> Success:
>> - State installation succeeded
>> - Reservation created
>> - Reservation accepted
> 
> Wouldn't an error bit set to zero in a response message
> indicate success? In that case, no parameter needs to be
> present. Or is this success indication meant to report
> some partial success only?

You could. My thinking was to prefer the consistency of having 'OK' as just another 'error' value. Otherwise you have the slightly more complicated "Is it an OK? (yes/no) If no, what is the error value?", but you do save space by not having to include an error object in the simple 'OK' case.

> [snip]
>> Permanent Failures:
>> - Resources pre-empted
>> - No NTLP reverse-path forwarding state
>> - NSLP soft-state expired
> 
> These errors may be transient, if transient simply
> means "try again later".

Yes, this area needs a bit more thinking about. What are the exact definitions of "permanent" and "transient" (and even "informational")?

Diameter doesn't particular help my understanding here. (Why is DIAMETER_AUTHENTICATION_REJECTED a transient failure and not a permanent failure?) For SMTP the retransmission mechanisms on a transient error are fairly clear (keep periodically retrying for a while and eventually give up).

I think you have to define the types by what you do in response. For transient errors in NSIS my current tentative definition might be something like "Transient failures indicate that the action could not be performed at the time it was received. If the same message is resent the action might be successfully completed without requiring any other protocol actions for this signalling session first." But that requires more rewording.

One of the problems is that almost any error could be transient, provided that the time period you're thinking about is long enough and you're allowed to do other things in between. :-)

regards,
Andrew

_______________________________________________
nsis mailing list
nsis at ietf.org
https://www1.ietf.org/mailman/listinfo/nsis