No support for any encoding other than 8-bit ASCII. In particular, no support for UTF-8 or UTF-16 (both required by the XML spec).
Not a validating parser, though it does check for well-formed documents.
Not all XML well-formedness checks required of a non-validating parser are made. These checks are being added. Here's the list of those that are missing. The values in parentheses are the internal document links for the HTML version of the XML spec.
External Subset (#ExtSubset)
PE Between Declarations (#PE-between-Decls) (Parameter entity replacement text must match the extSubsetDecl ::= ( markupdecl | conditionalSect | DeclSep) *, but NQXML does not yet recognize conditional sections).
PEs in Internal Subset (#wfc-PEinInternalSubset)
No External Entity References (in attribute values) (#NoExternalRefs)
Entity Declared (when standalone='yes') (#wf-entdeclared) (Currently all entities must be declared when references to them are parsed.)
No Recursion (#norecursion) (of entity references) Failing to enforce this constraint means that NQXML can enter an infinite loop.
No support for DTDs (though internal ENTITY tags are parsed). For example, attribute default values defined in the DOCTYPE are not yet used to modify attributes with missing values. Note that supplying default attribute values is not required for a non-validating processor such as NQXML.
Doesn't follow external references (therefore doesn't read ENTITY tags from an external DTDs). Note that following external references is not required for a non-validating processor.
ELEMENT, ATTLIST, and NOTATION tags simply slurp in their arguments and present them as strings. This will be fixed in a future release (but will probably break code that relies on the contents of the entities' :text attribute).