Default and empty CURIE prefixes in a non-XHTML host language

Dear RDFa developers,

  I am currently specifying the usage of RDFa inside a new host language
(OMDoc, a semantic markup language for mathematical documents;
http://omdoc.org).  I would like to simplify the explanations on CURIEs a bit,
compared to the RDFa syntax recommendation.  In OMDoc as a host language, we'd
like to
1. allow for using the XML default namespace for CURIEs in the default
   namespace, such as :name.
2. define our own vocabulary for CURIEs without prefix and colon.

I suppose both is possible, because the RDFa recommendation says that for
CURIE processing mappings for (1) and (2) have to be provided [by way of the
specification of the host language].  I think I will do that, but not
particularly endorse its usage, as general-purpose RDFa processors would not
understand it.

Now I'm just wondering why this is done in such a strange way for RDFa in
XHTML.  Why are the bare words that _are_ allowed in @rel and @rev
attributes not specified as prefixless CURIEs (mapping to the
http://www.w3.org/1999/xhtml/vocab# namespace) but as special reserved words?
Is this for historical reasons?  And why is there, in addition, a default
namespace http://www.w3.org/1999/xhtml/vocab# for use with the empty prefix?
That makes @rel="next" redundantly equal to @rel=":next".

There is additionally a separate CURIE spec at http://www.w3.org/TR/curie/.
Which one is more reliable, RDFa or CURIE?

Thanks in advance,

Christoph

-- 
Christoph Lange, Jacobs Univ. Bremen, http://kwarc.info/clange, Skype duke4701