Do we really need `VERSION` in N-Triples/Quads syntax?

[Tpt]

VERSION has been introduce in N-Triples for consistency with Turtle/TriG. However, I am unsure if it's a good call.

The "pro" I see is that it allows parser to choke early on the VERSION statement (or return a nice warning?). However:

1. This makes NTriples syntax more complex. One can't just "parse" the files into a 3 columns subject/object/table by filtering out empty lines, trimming the end of line . then splitting twice on whitespaces sequences.

2. I am not sure VERSION will get used in practice. Indeed, most NTriples files are programmatically generated (people tend to write Turtle by hand) and for compatibility reason we suggest that streaming serializers don't emit VERSION to avoid backward compatibility issues. And most of the NTriples serializers are streaming (rdf4j, N3JS, oxttl, Jena, RDF4J...). So, to be emitted, users would have likely to opt-in to actively set it in the file and I am not sure people will actually bother for just slightly nicer error from old parsers (I definitely wouldn't).

Hence, it seems to me the added complexity in the NTriples syntax is not worth it (NTriples aims at being simple and regular)

[afs]

On its own maybe that's not enough, but there is the use to record the version in the data for the long-term, such as archived data publications.

At 1.1 to 1.2, that isn't so important but VERSION is also some future proofing should a future RDF have a feature that is not so cleanly a syntax error if used.