The following is currently true in MEX:

1.) Metadata dialects/types are not required to define a "identifier". For example, there could be a dialect/type of {http://www.example.org}foo (to use the new serialization) that does not define an "identifier"; there is no single element or attribute value in a "foo" document that makes any different from any other "foo" document.

2.) @Identifier is an optional attribute of mex:PutMetadata/mex:Metadata/mex:MetadataSection.

3.) Services that implement MEX are not required to understand all the metadata dialects/types that they are asked to manage.

So what happens a client does a PutMetadata with one or more MetadataSections that have no @Identifier (whether this is because the contained metadata does not define an identifier or the client simply decided to omit the @Identifier, the service can't tell - see (3)), then someone else comes along and does a GetMetadata with @Dialect="{http://www.example.org}foo" and @Identifier="http://www.w3.org/2002/ws/ra/test/scenario"? Do hunks of metatdata that were Put with no identifier match against all identifier queries or none? There are use cases to support either answer to that question.