Re: Issue 298: RPC array representation unneccesary

Martin Gudgin wrote:

>I'm inclined to agree with the commentator[1] on this, the only
>difference between the struct and array case really is that duplicate
>edge labels are allowed in the array case. 
>

Also, in the struct case are nil items not omittable, but this would not 
make sense in the array case (it doesn't make sense to me in the struct 
case either except to complicate the receiver's job).

>I'm not convinced of the
>utility of that distinction and I think the spec would be clearer if we
>only speced the struct case. That said, I'd like to defer this issue for
>a week while Henrik and I get some more feedback from implementers
>inside Microsoft.
>  
>

It has recently been raised on soapbuilders that it is not possible for 
the receiver to know whether the sender intended the response to be 
interpreted as an array or struct representation in the case that there 
is no return value but only out values (in fact, even with a return 
value it may not be possible to tell without more metadata than WSDL 
provides). Since there is no way to say which representation the 
response contains there is a definite problem here for the receiver.

I don't think 2 distinct representations should be supported and would 
prefer settling on just the struct case (without omittable members!)..

Pete

Received on Saturday, 31 August 2002 20:00:40 UTC