On Thursday, Liam Quin wrote:
> numeric character references are always treated as data when they are read, not markup.
Note that's true but needs to be read with care as the same string may be read twice (which is why the definition of < requires double escaping).
<!DOCTYPE x [
<!ENTITY fooa "A <b>x</b>">
<!ENTITY foob "B <b>x</b>">
<!ENTITY fooc "C <b>x</b>">
]>
<x>&fooa; &foob; &fooc; </x>
foob using an entity reference < has replacement text using < so when foob is is referenced it generates the < as character data
but for fooa, the numeric character reference < produces the character < in the replacement text of the entity fooa, so when that entity is used the < starts a <b> start tag, so fooa is the same as fooc and produces a b element.
$ rxp -x lt.xml
<x>A <b>x</b> B <b>x</b> C <b>x</b> </x>
