Re: HTTP fault codes

My approach has been to adhere to the binding specific recommendations
and guidelines when such exist.  The <Result> element provides one way
to do this when there recommendations and guidelines do not exist.

<Result> is not abstract btw.

Regards,
Tommy

On 5/10/05, Frederic DELEON <frederic.deleon@crf.canon.fr> wrote:
> Hi,
> 
> I agree that <Result> abstract element could be used.
> But what is the SOAP fault interest if all errors can be managed through
> Result element through ResultMajor and ResultMinor ?
> Don't we have to keep same levels of errors between HTTP and SOAP bindings ?
> 
> I don't size up clearly the separation between XKMS errors (Major/Minor)
> and binding errors. For instance, what mechanism do I use when the
> server receives a request for an operation that is does not support ?
>   - SOAP fault with env:Sender/xkms:MessageNotSupported ?
>   - or XKMS result with Sender.MessageNotSupported ?
> 
> I have the same question for other cases.
> Don't you think there is an overlapping between SOAP fault 3,4 and 5 and
> XKMS result codes ?
>   - 3 : SOAP:Receiver could be managed through XKMS:Receiver.Failure
>   - 4 : SOAP:Sender.MessageNotSupported could be managed through
> XKMS:Sender.MessageNotSupported
>   - 5 : SOAP:Sender.BadMessage could be managed through XKMS:Sender.Failure
> 
> Regards,
> 
> Frederic
> 
> 
> Tommy Lindberg wrote:
> 
> > Hi Frederic -
> >
> > The beauty of the result messages is that their type hierarchy
> > includes a general <Result> element that can returned in response to
> > any request.  I use this in situations that you describe below, i.e.
> > if something goes terribly wrong and provided I catch all exceptions,
> > I can still return a valid XKMS result message; I could even sign it.
> >
> > Regards,
> > Tommy
> >
> > On 5/9/05, Frederic DELEON <frederic.deleon@crf.canon.fr> wrote:
> >
> >>Hello,
> >>
> >>In bindings specification, SOAP faults are described inside SOAP binding
> >>chapter.
> >>On the other hand, nothing is defined for HTTP binding, and for instance
> >>nothing is defined for HTTP faults. Are we free to implement this as we
> >>like ? Is it possible to return free XML error message inside HTTP
> >>response or do we have to return HTTP error such as 500 error code ?
> >>
> >>Frederic
> >>
> >>
> >>-- Frederic Deleon
> >>-- Canon
> >>
> >
> >>
> 
>

Received on Wednesday, 11 May 2005 22:53:34 UTC