Duplicate token in enumerated attribute declaration Duplicate token in NOTATION attribute declaration An unused attribute default need only be syntactically correct An attribute default must be syntactically correct even if unused Declarations mis-nested wrt parameter entities are just validity errors (but note that some parsers treat some such errors as fatal) Empty content can't contain an entity reference Empty content can't contain a comment Empty content can't contain a PI Empty content can't contain whitespace Element content can contain entity reference if replacement text is whitespace Element content can contain entity reference if replacement text is whitespace, even if it came from a character reference in the literal entity value Element content can't contain character reference to whitespace Element content can't contain entity reference if replacement text is character reference to whitespace Element content can contain a comment Element content can contain a PI Mixed content can contain a comment Mixed content can contain a PI External entity containing start of entity declaration is base URI for system identifier Parameter entities and character references are included-in-literal, but general entities are bypassed. Tokens, after normalization, must be separated by space, not other whitespace characters UTF-8 entities may start with a BOM Either the built-in entity or a character reference can be used to represent greater-than after two close-square-brackets Contains an irregular UTF-8 sequence (i.e. a surrogate pair) Three-letter language codes are allowed A non-deterministic content model is an error even if the element type is not used. An external ATTLIST declaration does not make a document non-standalone if the normalization would have been the same without the declaration XML 1.0 document refers to 1.1 entity An xml:lang attribute may be empty ANY content allows character data All line-ends are normalized, even those not passed to the application. NB this can only be tested effectively in XML 1.1, since CR is in the S production; in 1.1 we can use NEL which isn't. A reference to an unparsed entity in an entity value is an error rather than forbidden (unless the entity is referenced, of course) A value other than preserve or default for xml:space is an error Conditional sections are allowed in external parameter entities referred to from the internal subset. (From John Cowan) An encoding declaration in ASCII specifying an encoding that is not compatible with ASCII (so the document is not in its declared encoding). It should generate a fatal error.