Major Description
N-Triples is a subset of Turtle, and it is also a subset of N-Quads. Turtle is a subset of TriG.
But when we are performing mimetype detection on a plain text file, we only sniff the first few kilobytes of data. Therefore, something we initially detect as N-Triples may in fact be a Turtle, Trig, or NQuads document. Something we initially detect as Turtle may in fact be a TriG document.
Therefore, if we detect that the document is Turtle, in the absence of a declared Content-Type, we should probably assume that it actually TriG, just in case.
If we can only detect that the document is N-Triples, that presents a problem, because it could also be either Turtle or N-Quads. Which do we choose?
Another problem I see is that we are detecting both N3 and Turtle in two separate steps. However, as I understand it, for the purposes of RDF, N3 is essentially a synonym for Turtle. So it doesn't really make sense to use two different detection steps for this. It appears that our N3 detection step is actually detecting N-Triples, which is not at all the same thing.
(Indeed, in org.eclipse.rdf4j.rio.n3.N3ParserFactory's implementation of getParser() we see: return new TurtleParser())