SoTA research?

[berezovskyi]

Did the group do a research on the state of the art before zeroing on CBOR and the direction of the spec? Examples:

• https://jelly-rdf.github.io/dev/specification/serialization/
• https://jena.apache.org/documentation/io/rdf-binary.html

[Ostrzyciel]

I did a quick experiment comparing CBOR-LD and Jelly a while back: https://ostrzyciel.eu/blog/2025/jelly-and-cbor-ld/

The short version is that these formats have very different use cases and trade-offs – same goes for HDT. In the tabular world, you also have Arrow, Parquet, Avro, Iceberg, etc, etc. It all depends on your use case. I think it would be useful to consider first what are the design goals for CBOR-LD – you can't optimize for everything.

Also to consider:

• https://rdf4j.org/documentation/reference/rdf4j-binary/ (the spec is out of date)
• https://openengiadina.codeberg.page/rdf-cbor/ (apparently the creator is now inactive due to political oppression...)
• And of course HDT

[davidlehn]

I don't recall what was looked at when this project was started. Obviously if we're calling it CBOR-LD it should use CBOR. :-) The algorithmic techniques here could likely be applied to other projects. Similarly, we should look at what other projects are doing and see what could be applied here. I think there's room for experimentation and exploration and multiple projects in this space. There are going to be tradeoffs based on use cases and the tech used. And we might not know what works well until it's built and tested.

From an information view, I would guess any system would have similar maximal compression depending on what algorithms are used. It's probably maxed out when you externalize everything possible based on a specific pre-known use case. CBOR-LD is pushing in that direction while also being able to handle arbitrary JSON-LD data.