Small bug in XDI address ABNF

[Drummond Reed <drummond.reed@xdi.org>]

Markus gets the 2010 New Year's prize for noticing a small bug in our four-line ABNF for XDI addressing (below) while working on his new XDI Grapher application. 
xdi-address     = xdi-subject [ "/" xdi-predicate [ "/"  xdi-object ] ]

xdi-subject     = xri-segment
xdi-predicate   = xri-segment
xdi-object      = xri-segment
                / "/" [ xdi-address ]
The problem is that our ABNF assumed that every XRI segment was valid as a component of an XDI address. Markus pointed out that...
(+a/+b/+c/+d)
...is a valid XRI segment, but the XRI within this cross-reference is not a valid XDI address. Since XDI uses cross-references for reification, the contents of every cross-reference must itself be either a valid XDI address or a URI.

The fix is pretty easy - it just involves restricting xrefs to the rule above. The ABNF goes from 4 lines to 12. The following has been posted to the XDI RDF Graph Model spec wiki page.
xdi-address     = xdi-subject [ "/" xdi-predicate [ "/"  xdi-object ] ]
xdi-subject     = xdi-segment
xdi-predicate   = xdi-segment
xdi-object      = xdi-segment

                / "/" [ xdi-address ]
xdi-segment     = [ literal ] *xdi-subseg
xdi-subseg      = global-subseg
                  / local-subseg
                  / xref
global-subseg   = gcs-char [ local-subseg / xdi-ref / literal ]

local-subseg    = lcs-char [ xdi-ref / literal ]
xdi-ref         = "(" [ xdi-ref-value ] ")"
xdi-ref-value   = xdi-address
                  / iri
=Drummond