rdf4j and IPFS

[naturzukunft]

has anyone thought about connecting rdf4j and IPFS?

[kenwenzel]

This is an interesting idea. What kind of data would you like to integrate via IPFS?
I could image to query or create distributed HDT or Parquet files that contain partial RDF graphs.

[naturzukunft]

I couldn't really sort it out in my head yet ;-)
I'm toying with the idea of recording recipes and food data via ActivityPub. And it's a problem to distribute this as LinkedData, because then it's either centralised or there's a lot of redundancy, as the same data would be available under different URLs. Access via CDI would be very attractive. However, I see challenges in querying the data. That's why I'm currently considering a mixture of TripleStore and IPFS. But I would be happy to hear other views and ideas. Somehow there doesn't seem to be a technical solution for my vision yet.

[kenwenzel]

It also depends on the types of queries you like to run on the data. Do you need to search for data setd based on the contained values or is it more like retrieving one or more data sets by id and then running a query on them?

[naturzukunft]

1. search for food that is available on a certain date in a certain region.
2. search for recipes for these (1) foods

[Ostrzyciel]

The use case sounds a lot like nanopublications... https://nanopub.net
CC-ing @tkuhn who leads the tech development of this. Nanopubs are published to a decentralized network of registries that have "classic" SPARQL query endpoints attached to them. The endpoints replicate a part of the network's contents in a triple store. Currently this actually uses the RDF4J LMDB store.

Comunica is a very different approach, also partially matching what I think you are looking for: https://comunica.dev/
In their paradigm, data storage is distributed and the query engine runs your browser. This can work for some use cases, but if the data is not indexed by the value you are looking for using links... well, it's going to have to visit all those pages, which will be very slow.