[relaxng-user] Ambiguity in RNC grammar

[Bob Foster]

Jeff Rafter wrote:

> I have just about finished the translator I am working on for RNC->RNG. I
> stumbled upon what I think is an ambiguity in the grammar. Because XML
> allows for a "-" in a name, and RNC allows for "-" as part of a dataExcept
> (or nameExcept)-- I think that there is a collision. Consider:

You are expected to resolve this ambiguity during tokenization (see 
http://www.oasis-open.org/committees/relax-ng/compact-20021121.html#d0e4256). 
When the longest initial subsequence of an NCName contains a '-' 
character, it is part of the name. Therefore, a sequence like:

xsd:int-xsd:byte

consists of three tokens, a CName 'xsd:int-xsd', an invalid token ':' 
and an NCName 'byte'. Jing must check the datatype name before scanning 
the next token, or you would get a syntax error. Other processors might 
do it in reverse order.

Bob Foster