files should be saved to disk using a predictable Turtle Writer

[bblfish]

( or predictable JSON-LD writer! whatever! )

We are saving our data to disk to make it readable and easy to interact with. But if every time we save it in a randomly different manner ( as it is now), then saving to disk makes that feature a complete non feature: no human will be able to interact with a randomly changing rdf file like that. Recent talk on the semantic web mailing list has brought back talk on stable bnodes ( for PATCH ) that make it easy to create such stable serialisations.