Test issues

A summary of some of the contentious tests

elementreplaceattributewithself:  

Seems to follow the letter of the spec, but only Mozilla do far passes the test and that is probably unintentional since it doesn't throw IN_USE_ERR in places that it should.  If "attached to another element" was changed to "attached to any" by an errata the existing behaviors would be conformant.  Otherwise, the test seems legitimate to me.

strict equality testing on URI's in entitygetpublicid and notationgetsystemid:

Only Crimson changes the relative URI's in the document to absolute URI.  Though not explicitly prohibited, it is unexpected and undesirable.  Possible resolutions are: 

a) requiring the processor to return precisely the URI from the document

b) an errata warning the user to anticipate that a processor might absolutize an URI.  

c) some accomodation of existing Crimson behavior with URI preservation for L3+


*EE tests and documentcreateentitydefaultvalue:

These test check that create entity references have the entity content available.  MSXML and Xerces-J 1.4.3 in entity preserving mode pass these tests.  Xerces-J 1.4.3 in entity expanding mode and Crimson fail the tests