From: Bjoern Hoehrmann <derhoermi@gmx.net> 

The lack of clearly identified normative references and the lack of a
conformance section introduce a lot of ambiguity, it is for example not
clear what the document considers a NCName. Is that an NCName as defined
in XML Namespaces 1.0 or as defined in XML Namespaces 1.1? Maybe it
depends on whether the document uses XML 1.0 or XML 1.1? Is it at all
allowed to use XML 1.1 for CDL documents?

Another example is XInclude, the document notes

[...]
  A Choreography Package aggregates a set of WS-CDL type definitions,
  provides a namespace for the definitions and through the use of
  XInclude [XInclude], MAY syntactically include WS-CDL type definitions
  that are defined in other Choreography Packages.
[...]

in section 2.2.1, but that's already clear from section 2.2.4 which
notes

[...]
  To support extending the WS-CDL language, this specification allows
  the use of extensibility elements and/or attributes defined in other
  XML namespaces. Extensibility elements and/or attributes MUST use an
  XML namespace different from that of WS-CDL. All extension namespaces
  used in a WS-CDL document MUST be declared.
[...]

So maybe the former text means that processors must support XInclude?

Another example is section 2.2.2 which notes

[...]
  A WS-CDL processor MUST ensure that the document is correct before
  processing it. The correctness may involve XML well-formedness as well
  as semantic ;checks, such as unicity of Variable definitions, of a
  single root Choreography, etc.
[...]

Well, it needs to be clearly specified what correctness involves, how
else should it be possible to implement this requirement interoperably? 

In summary, I think there is not much point in issuing a Last Call
announcement with integral parts of the specification such as the
conformance section missing.