How to store RDF data on the SAFE Network

[bochaco]

@intrz

I don’t really see much reason to actually actually serialiaze to and storing Turtle, JSON-LD etc if you can get the data in those formats with an API anyway.

I’m hopefully not getting you wrong, but this is the same initial thought I had, then I started thinking how can we store the triples in the key value store, so let’s say I do what you suggested:

@intrz

To use this model in a key value store you could try something like save a triple with the subject as the key an predicate,object as the value, predicate as the key and subject,object as the value and object as the key and predicate,subject as the value. Then maybe add in a graph name in the key for namespacing purposes.

What you are doing here is basically creating a new type of text based serialisation, because you need to serialise the concatenation of predicate and object with a , or a |, to store it as the entry’s value, right?
Then to read the key and value you’ll need to have a parser for that serialisation you just defined, a very simple one perhaps, but this is still a serialisation format you define, which is not compressed as e.g. Turtle or JSON-LD grouping triples with the same subject (graphs), and which is not an standard one, so you are not really doing what we think we want or desire to have:

@bochaco

3. We want to store the RDF using any of the standard serialisation formats

In fact this is exactly what I impulsively did for a few minutes when started looking into this, I literally sat down and started writing down some ideas of how to put the subject, predicate and object in the MD entries, and I then realised that I was trying to come up with my own serialisation format, so why do I need one more if we already have several (no even to mention the discussions I’d have to go thru to convince people it’s better than the existing standard ones, you can review Manu Sporny experience with something like that )

In addition to this, I should have mentioned that having a triplestore implemented is something that is also being considered by some of us, but we don’t know how crazy it is, and we have to evaluate pros/cons of it, it sounds like it could make sense, in that way you can store the triples in let’s say a “raw form”, and you can even apply indexing mechanisms when storing the triples as you described. BTW, thanks for those ideas and links, I’ll be looking at them, I hope I’m making sense.

EDIT: Just in case it helps, there is an older discussion related to all this as well, that we had before we implemented the PoC: Resource Description Framework implementations and 'Portable Data...': Some Links and learnings