The purpose of the RDF stack is to give applications access to a pervasive, abstract, per-user semantic space (see AndersFeder/SemanticSpace).
The central pillar of the RDF stack is the RDF store. The RDF store consists of a (SPARQL) query parser and support for one or more graph backends. The store should accept queries through a D-Bus API. In the early stages, Soprano can act as this store. Later on, a GNOME-native (e.g. Redland-based) store may be implemented.
The RDF store interface with backends through a D-Bus API. This D-Bus interface is to be defined and implemented in Soprano.
A library for implementing the backend API easily will be made available.
As a proof-of-concept, an Evolution data server backend will be designed, using this library. (This should be considerably easier now in light of Project Madagascar.)
An interim, generic XESAM backend will also be made, such that e.g. Tracker's database can be queried through the stack.
-- MikkelKamstrup 2008-06-23 20:30:33 : Regarding Xesam let me point you at the query parser in xesam-glib. It is designed to be a "translation layer", so using it to generate sparql from a Xesam query (or query in the user search language) should be straight forward. This query parser is included from 0.3.0 and forward, 0.3.0 will be released tomorrow.
AndersFeder: Thanks Mikkel.
A service for persistent queries will be available.
Finally, a library for managing resources in the semantic space as GObjects, tentatively called libgsemantic, will also be available (see AndersFeder/libgsemantic).