4 ms·

It is really great that people are discovering the utility of triplestores through their ubiquity in the Clojure ecosystem (e.g. Datomic and its many clones).

It is really great that people are discovering the utility of triplestores through their ubiquity in the Clojure ecosystem (e.g. Datomic and its many clones). I do feel saddened that few seem to realise that the fundamental idea actually comes from RDF (i.e. the semantic web) and that Datomic's Datalog dialect is actually very similar to SPARQL. Datomic's primary innovation is the immutability/time travel feature, not the fact that is uses triples and works like a graph database. If you find triplestores fascinating I implore you to explore the semantic web stack too.

macmac 4y ago
Rich often mentions the RDF heritage. I think the separation of reads and writes through the transactor is also a significant innovation.
j-pb 4y ago
The most important thing that datomic and friends brought, is usability and pragmatism. While RDF has an elegant core, it is bogged down by the complixity of the web and ancillary technologies. XML Types, cURIes, OWL, the different encodings. They all make RDF a nightmare to work with, and to implement in practice. The is simply no good RDF software ecosystem, yet the Spec ecosystem continues to grow (wildly out of control). On a technical note, SPARQL and Datomic Datalog are very different beasts. The former is conjunctive queries with optionals and regular path expressions (i.e. Conjunctive queries + XPath + nonmonotonic defaults) while the latter is recursive conjunctive queries (i.e. Datalog).
greendream17 4y ago
If PostgreSQL is to SQL what the browser is to HTML then what is X to SPARQL. One of the reasons why I don't think X "exists" is because it's not a good way of working with graph. I sometimes think there should be some alternative tech stack to RDF/SPARQL that isn't part of the semantic web but uses many of the same ideas but you know something that people can use.
- amanj41 4y ago
  I suppose X in this case would be any of the number of graph databases out there. Ontotext GraphDB, Virtuoso, AWS Neptune. Of course there are other GraphDBs with additional properties and custom query languages such as Neo4j and its query language Cypher. I Find that Neo4j's property graph is more intuitive to model data with than a plain RDF store and Cypher has some advantages over SPARQL as well
cscheid 4y ago
I'm sorry, but Datalog has existed since 1977; RDF, since 1999.
- simongray 4y ago
  Good thing I was talking triplestores and not Datalog then.

macmac 4y ago

Rich often mentions the RDF heritage. I think the separation of reads and writes through the transactor is also a significant innovation.

j-pb 4y ago

The most important thing that datomic and friends brought, is usability and pragmatism. While RDF has an elegant core, it is bogged down by the complixity of the web and ancillary technologies. XML Types, cURIes, OWL, the different encodings. They all make RDF a nightmare to work with, and to implement in practice. The is simply no good RDF software ecosystem, yet the Spec ecosystem continues to grow (wildly out of control). On a technical note, SPARQL and Datomic Datalog are very different beasts. The former is conjunctive queries with optionals and regular path expressions (i.e. Conjunctive queries + XPath + nonmonotonic defaults) while the latter is recursive conjunctive queries (i.e. Datalog).

greendream17 4y ago

If PostgreSQL is to SQL what the browser is to HTML then what is X to SPARQL. One of the reasons why I don't think X "exists" is because it's not a good way of working with graph. I sometimes think there should be some alternative tech stack to RDF/SPARQL that isn't part of the semantic web but uses many of the same ideas but you know something that people can use.

amanj41 4y ago
I suppose X in this case would be any of the number of graph databases out there. Ontotext GraphDB, Virtuoso, AWS Neptune. Of course there are other GraphDBs with additional properties and custom query languages such as Neo4j and its query language Cypher. I Find that Neo4j's property graph is more intuitive to model data with than a plain RDF store and Cypher has some advantages over SPARQL as well

cscheid 4y ago

I'm sorry, but Datalog has existed since 1977; RDF, since 1999.

simongray 4y ago
Good thing I was talking triplestores and not Datalog then.