a general question:why RDF?

Dear All ,
I have been developing lightweight   owl graphic browser/editor in form
of a java applet:
http://ecoinformatics.uvm.edu/dmaps/growl/ 
However in my first encounter with parsing/rendering rdf/owl I found
that it is really difficult to find a lightweight solution.  The
slimmest Wonderweb's owlapi had to be trimmed to produce 600k  applet.
My previous experience with XML based specification for a language with
similar complexity where far easier. Besides being frustrated with
complexity of RDF processing I completely do not understand why RDF is
necessary as an intermediate layer between XML and OWL. Why simple XML
solution http://www.w3.org/TR/owl-xmlsyntax/ can't be adopted as a
standard.
 
As I understand it after [1] and other papers written by logicians  RDF
on its own does not have any discernable  advantages   in terms of
expressive power , decidability, (or even completeness ??). It did not
have even formal semantic when it was adopted by semantic web. The non
standard semantic it has now seems does not correspond to the semantic
of owl. And it looks like RDF can be used for (clean) reasoning only
with some 'well behaved' languages on the top. Then, after adopting OWL
as a standard, what is the added  value of RDF in comparison to XML? 
 
May be I am merely missing  some points and someone can explain me?
 
Thanks.
Serguei Krivov, Ph.D. ,Computer Science Dept & Gund Institute for
Ecological Economics, University of Vermont; 590 Main St. Burlington VT
05405 phone: (802)-656-2978
 
 
 
[1]   Ian Horrocks and Peter F. Patel-Schneider. Three theses of
representation in the semantic web. In Proc. of the Twelfth
International World Wide Web Conference (WWW 2003), pages 39-47. ACM,
2003.
http://www.cs.man.ac.uk/~horrocks/Publications/download/2003/p50-horrock
s.pdf