The features of WS-ResourceTransfer are a natural extension of WS-Transfer 
and really should be part of the WS-Transfer spec itself.  The reason for 
the current split is mainly history. Now would be the time to merge the 
fetures of RT into T so that we remove the need for an additional 
specification and reduce the complexity for the implementators and end 
users.

Proposal:
Move the following features from RT in T:
- Fragment support
  Pull would then become:
  <wst:Get Dialect="xs:anyURI"?>
    <wst:Expression ...> xs:any </wst:Expression> *
  </wst:Get>
  Note: if Dialect is not present then wst:Get acts like it does today.
  In other words, the child of wst:Get becomes the full representation of 
the resource.
 
  PullResponse would then becomes:
  <wst:GetResponse>
    <wst:Result ...> xs:any </wst:Result> *
  </wst:GetResponse>

  <wst:Put Dialect="xs:anyURI"?>
    <wst:Fragment Mode="Modify|Insert|Remove">
      <wst:Expression>xs:any</wst:Expression> ?
      <wst:Value ...>xs:any</wst:Value> ?
    </wst:Fragment> +
  </wst:Put>
  Note: if Dialect is not present then wst:Put acts like it does today.
  In other words, the child of wst:Put would becomes the full 
representation of the resource.

  Create would become:
  <wst:Create Dialect="xs:anyURI"?>
    <wst:Fragment>
      <wsmex:Metadata> ... </wsmex:Metadata> ?
      <wst:Expression>xs:any</wst:Expression> ?
      <wst:Value ...>xs:any</wst:Value>
    </wst:Fragment> *
  </wst:Create>
  Note: if Dialect is not present then wst:Put acts like it does today.
  In other words, the child of wst:Create would becomes the full 
representation of the resource.

- Move over the Dialects
- Move all faults associated with these features.

Note: if we do this then some of the other issues I've opened about BP 
compliant will go away because this already adds a wrapper/GED to the 
child of the env:Body.