Right now there is a issue for getting token into core and one for getting some RDF into it. I wonder if makes sense to base the token patch upon the RDF stuff. So once we have some semantics attached to our objects, we can use them as a starting point for doing the token replacements.
The other way round, the RDF patch could learn from the tokens code - as the actual problem to solve is quite similar. The way the patch in #74 uses simple handlers to get tokens would make sense for RDF too. Then the token implementation can go and build upon that and reuse the handlers returning RDF string literals. Similarly, the approach described in #77 for doing actual token replacements could also be used for getting chunks of RDF for a bunch of objects.
Then apart from avoiding inventing the wheel twice, we could easily make both APIs consistent, so they can be used the same way.