Hi, I'd like to point you to a project of mine that might be useful to Jena users. Jelly is a high-performance binary format and streaming protocol for RDF based on Protobuf. It was designed to be extremely fast in serialization/deserialization, while also offering decent compression (about 1/6 of N-Triples and 1/3 of Turtle). In the benchmarks that I ran <https://w3id.org/jelly/dev/performance/>, it was faster than all other serializations in Apache Jena, including RDF Thrift and RDF Protobuf. Of course, this is a first-party benchmark, so take it with a grain of salt. :)
Jelly-JVM <https://w3id.org/jelly/jelly-jvm/> is an Apache 2.0-licensed implementation of Jelly that integrates with Jena (it works with RDF4J as well). You can use it as either a drop-in JAR plugin <https://w3id.org/jelly/jelly-jvm/dev/getting-started-plugins/> to quickly add Jelly support to Fuseki and CLI tools of Jena, or as a Maven dependency. Jelly-JVM 1.0.x supports Jena 4, while Jelly-JVM 2.0.x supports only Jena 5. The two versions of the library are near-identical otherwise (same binary format, same compression algorithm). Jelly was specifically designed to work with streams, so if your application involves streaming RDF over gRPC, Kafka, or similar, then Jelly may be able offer you much better performance than traditional serializations. It should also be faster for plain old files as well, though. Any questions, feature suggestions, pull requests, or bug reports are of course very much welcome. :) I'm very curious if you will find this project useful and if you see any possible areas of improvement. The issue tracker is here: https://github.com/Jelly-RDF/jelly-jvm/issues If you want to try out the Jena JAR plugin with Jelly, you can follow the installation instructions here <https://w3id.org/jelly/jelly-jvm/dev/getting-started-plugins/> and try loading any of the datasets listed on the bottom of this page <https://w3id.org/riverbench/v/dev/profiles/flat-mixed-rdfstar#jelly-distributions>. They should "just work" by dropping the downloaded file into Fuseki UI, or by using the usual CLI utilities of Jena. You can also download the equivalent datasets in N-Triples for comparison. Please reach out to me in case of any questions. And thanks to Andy for suggesting posting this to the mailing list. -- Piotr Sowiński / Ostrzyciel
