In my last post about the ARC parser, I said that RAP failed on parsing non-ASCII rdf:IDs. While doing some benchmarks today, I found a simple option to allow extended ID values, so RAP *does* pass tests with extended characters. And the XML Literal issues seem to be related to the way they included my unicode NFC function: RAP by default converts every subject, predicate, and object to the normal form C, although this would only be required for serializing a model as N-Triples. This leads to an unneccessary (huge) performance loss (the time for parsing the planet rdf feed went down from 11 seconds to 3 after turning off the auto-encoding).

I'll send a mail to the RAP team in a second, but in case you don't want to wait for a fix, simply set the 'UNIC_RDF' constant (in api/constants.php) to false and your parser should work fine again.

I guess that running the string converter during the parsing *and* the serializing process is the reason why RAP's XML Literal output was no longer accepted by my test case walking script. Should be easy to fix then.