- From: Jonathan Marsh <jmarsh@microsoft.com>
- Date: Fri, 6 Sep 2002 11:50:49 -0700
- To: "Joyce Yang" <joyce.yang@oracle.com>, "Gaertner, Dietmar" <Dietmar.Gaertner@softwareag.com>
- Cc: <www-ws-desc@w3.org>
A solution to the base problem might be to restrict the encodingStyle attribute value to an absolute URI. > -----Original Message----- > From: Joyce Yang [mailto:joyce.yang@oracle.com] > Sent: Thursday, September 05, 2002 3:14 PM > To: Gaertner, Dietmar > Cc: 'www-ws-desc@w3.org' > Subject: Re: New issue: Can the "use" attribute be eliminated? > > > A few comments on the proposal -- > > 1) encodingStyle="..." does *not* mean that the use is encoded (section > 3.5 in WSDL 1.1 and section 2.5 in WSDL 1.2 binding spec). One > can have encodingStyle for doc/literal. This is the case of "writer makes > right" where there is a schema/algorithm to map a particular data model > to XML data model and the encodingStyle provides a hint to the receiver > as to how the receiver should map the XML data back to the application > specific data model. > > 3) using empty string value to indicate literal is problematic. If XML > Base is in scope then an empty string is a relative URI wrt to XML Base. > > -Joyce > > "Gaertner, Dietmar" wrote: > > > I took an action item in today's telcon to formulate > > a possible new issue on whether the "use" attribute > > isn't redudant and can be eliminated. This has been > > discussed (among others) in the soaptf [1]. Following > > is the "use" attribute rationale extracted and slightly > > re-formulated. > > > > Can the "use" attribute be eliminated? > > -------------------------------------- > > > > The "use" attribute (soap:body, soap:header and soap:headerfault > element) > > has possible values: "literal" and "encoded". The following combinations > > of style/use are possible: > > document/literal - makes sense > > document/encoded - makes sense (e.g. for docs using the SOAP data > model) > > rpc/literal - does this make sense? Probably not, because RPC > > implies a special encoding or format. > > rpc/encoded - makes sense (RPC even requires encoded) > > > > Given that there can be used different encoding styles, and when > > we have use="encoded" also encodingStyle="..." has to be specified, > > isn't use="encoded" redundant? Even more, > > - doesn't encodingStyle="someURI" imply use="encoded" > > - and isn't a missing encodingStyle or encodingStyle="" equivalent > > to use="literal"? > > > > Proposal: > > As the "use" attribute appears to be redundant > > eliminate it and just use the "encodingStyle" attribute > > to express "literal" via an empty string value and "encoded" > > via a non-empty string value. > > > > See also: > > Issue 45 [2] and 48 [3]; "use" attribute of [...] should be optional > > and Arthur's encodingStyle proposal [4]. > > > > [1] http://lists.w3.org/Archives/Public/www-ws-desc/2002Jul/0039.html > > [2] > > http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/issues/wsd- > issues.html#x45 > > [3] > > http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/issues/wsd- > issues.html#x48 > > [4] http://lists.w3.org/Archives/Public/www-ws-desc/2002Sep/0018.html > > > > Regards, Dietmar.
Received on Friday, 6 September 2002 14:51:21 UTC