Copying some notes from a discussion with @ianconsolata (Filecoin Foundation) here:
- Are there good intro to DASL or migrating to DASL guides I can send to the Singularity PM so we can make sure Singularity is packing CARs in a DASL compatible way?
- We should also chat about inline prepared CARs: Inline Preparation | Singularity They were useful for IA, because they allow pointing to the existing data on disk rather than duplicating storage with a full copy of the data in a car file. But I suspect that approach is probably incompatible with the current DASL approach to CARs.
- I think DASL (and IPFS) are agnostic about how the CARs are generated. But in practice, the current CAR generation tools are impractical to use for large files/datasets because they require 2x storage to generate?
- Right — it’s a tooling issue, we could definitely help have better tooling.