The w3c published a new recommendation for entity names: http://www.w3.org/TR/xml-entity-names/. MathML 3 should implement all these entities and HTML 5 only a subset. However, the actual entity parser cannot handle these new entities. Thus, the entity parser should be rewritten in the following way (see discussion on the mailing list): - use a prefix tree instead of a hash table (faster and handle entity names as long as wanted) - handle entities associated to combined Unicode characters - solve the conflict with XHTML 1.0 about ⟩ and ⟨ (see w3c webpage above)