Pursuant to https://fedora-commons.org/jira/browse/FCREPO-400
There are four places in a service deployment object that a dataStream ID/wsdl message part name are indicated:
foxml:digitalObject / foxml:datastream[@ID="WSDL"] / foxml:datastreamVersion / foxml:xmlContent / wsdl:definitions / wsdl:message / wsdl:part@name
- Conditionality: Unclear
foxml:digitalObject / foxml:datastream[@ID="METHODMAP"] / foxml:datastreamVersion / foxml:xmlContent / fmm:MethodMap / fmm:Method / fmm:DatastreamInputParm@parmName
- Conditionality: apparently indicated by @required, although this element doesn't appear to be part of the assembly of dissemination URL's or sdep validation
foxml:digitalObject / foxml:datastream[@ID="DSINPUTSPEC"] / foxml:datastreamVersion / foxml:xmlContent / fbs:DSInputSpec / fbs:DSInput@wsdlMsgPartName
- Conditonality: Required if present, although @DSMin could imply a lack of requirement
- Note: This attribute value must match a DataStream ID in the relevant Digital Object
foxml:digitalObject / foxml:datastream[@ID="WSDL"] / foxml:datastreamVersion / foxml:xmlContent / wsdl:definitions / wsdl:binding / wsdl:operation / http:operation@location
- Conditionality: Required if present
- Note: This reference is solely in the form of URL substitution patterns
In the 3.1 release of DefaultAccess, the assembly of a dissemination URL:
- Parses the DSInputSpec section, and throws an exception if any DSInputs have a @wsdlMessagePartName that does not correspond to a DataStream ID
- Uses the @wsdlMessagePartName value in parentheses to establish a substitution pattern
- Replaces all occurences of the substitution pattern with the URL_REF value(s) indicated in the DataStream IDs
- Throws an exception if the resulting URL contains any more apparent substitution patterns in parentheses
It seems like it would be straightforward enough to write an Access module that simply didn't throw the exceptions in step 1 above, but that approach still seems to ignore a lot of information that may be encoded in the sdep. There is also some obscurity in the lack of a separation of the wsdl message part name and a bound datastream ID.
Maybe the responsibility could be divided up as:
- DSInput binds a datastream ID to the to a partName, perhaps by the addition of a datastreamId attribute
- MethodMap indicates that a named parm is taken from a datastream
- WSDL Message Part name is used to establish the substitution pattern
This makes some sense to me, but leaves open the issue of when to assert a missing part error, and also where to indicate optionality of a part (and what to do if an optional part is missing).