- From: Henrik Frystyk Nielsen <henrikn@microsoft.com>
- Date: Mon, 15 Apr 2002 15:56:31 -0700
- To: "Noah Mendelsohn/Cambridge/IBM" <noah_mendelsohn@us.ibm.com>, "Jacek Kopecky" <jacek@systinet.com>
- Cc: "Ray Whitmer" <rayw@netscape.com>, <xml-dist-app@w3.org>
FWIW (and take this with a grain of salt). I can live quite happily
without support for positional [in][out] parameters but just as a
proposal - if we want to go there then maybe we can get around the
current problem by saying that we *always* use a struct with or without
a "result" (status quo) and then model positional parameters as an array
within the struct:
...
<s:Body s:encodingStyle="http://www.w3.org/2001/12/soap-encoding">
<m:MyResponse xmlns:m="http://www.example.org/foo/bar">
<r:result
xmlns:r="http://www.w3.org/2001/12/soap-rpc">...</r:result>
<m:MyParams e:itemType="xs:int">
<item>1</item>
<item>56</item>
</m:MyParams>
</m:MyResponse>
</s:Body>
...
Does this make sense - am I missing something obvious?
Henrik
>Is it intentional that when modeling a function with positional [out]
>arguments, we cannot reliably determine whether the return
>value is void
>unless we have external knowledge of the method signature and
>the number
>of arguments expected? For a function taking a variable number of
>arguments, it would seem to be impossible to determine in
>general whether
>the return value was void.
Received on Monday, 15 April 2002 18:56:42 UTC