Re: Issue 192 & R803

Henrik Frystyk Nielsen wrote:

> Hmm, from an architectural point of view, I am somewhat uncomfortable
> make a fault special in this regard - it seems to break orthogonality
> between the envelope and faults. IMO, even though we in part 1 define a
> SOAP fault as the only "message-type", processing-wise the SOAP fault is
> separate from the envelope in that it defines its own semantics (what
> does "faultcode" mean etc.)
> 

I don't understand why making the fault a child of the envelope instead 
of the body breaks orthogonality with the envelope or changes the 
processing model - could you elucidate further ?


> From a practical point of view, it also seems to make the description of
> the envelope more complicated as it would mean that we can't talk about
> the body anymore as a unique thing. I think we already have the
> possibility for carrying SOAP fault EII even though they may not "count"
> as faults because a SOAP fault is *only* a SOAP fault in the processing
> sense *if* it is located as the first child EII of the body EII.
> 

In the spec we don't say anything about the fault having to be the first 
child EII of the body, only that it must be a direct child and that 
there should only be one fault EII.

We don't disallow other EIIs within the body along with a fault and we 
don't say anything about processing the fault or any EIIs that may 
accompany it.

Regards,
Marc.

-- 
Marc Hadley <marc.hadley@sun.com>
XML Technology Centre, Sun Microsystems.

Received on Thursday, 4 April 2002 05:43:54 UTC