IPFS Thing 2023: Data Transfer batching Techniques featuring Blake3, CAR Mirror, and more

Presentation by @matheus23 in the Data Transfer track.

Slides: https://bafybeih45t7oeqzlplns3v52zbl4opc73mfjhjhr2hnnkfdvdxrym43tji.ipfs.dweb.link/?filename=DataTransferBatchingTechiques.pdf

Comparison of Approaches

1 Like

Perhaps worth noting that :x: on CAR Mirror is the current implementation. It can (and probably will!) evolve to include a query language, but we wanted to ship incrementally.

Yep. I mention this a bunch in the talk itself: In theory both car mirror and graphsync can evolve to solve all design goals listed, and blake3+bao can gain actual inner deduplication, even if it’s much less useful without DAG layout flexibility.