-
Notifications
You must be signed in to change notification settings - Fork 14
Define Data Availability Sampling (DAS) #110
Comments
I'm not sure the number of samples has to be a part of the spec because clients can decide this for themselves, however the spec can make secure recommendations. |
Just for my understanding: will this include how we use ipfs? |
Eventually, yes. But it will probably start with more conservative goals, e.g. just defining what to sample. Then move on to how to sample as that gets iterated on in the implementation. |
A recent topic that came up is "do we need to sample from both rows and columns for DAS?" (cf. celestiaorg/celestia-core#270) Mechanism 1 (from paper)The mechanism from the paper (celestiaorg/celestia-core#270 (comment)) involves sampling shares randomly from both rows and columns. In other words, each share in the EDS (extended data square) has two possible addresses: the pairs A fraud proof consists of Merkle proofs against a single row/column root for at least Mechanism 2 (potential alternative)We note that having two addresses for shares introduces duplication without any additional security. Instead, we can only sample from rows. The space of shares is identical to above, so each individual share has exactly the same probability of being sampled. The only difference is that only Merkle proofs against row roots are downloaded. A fraud proof consists of Merkle proofs against a single row root (for a fraudulent row), or a Merkle proof against multiple row roots (for a fraudulent column) for at least DiscussionPros of alternative:
Cons of alternative:
I don't immediately see any security issues with the alternative, but maybe I'm missing something. |
The question is how much larger in the worst-case and how often do we expect this to happen (and who is impacted by this/who needs to download and process these fraud proofs)? |
Without deduplication of sibling nodes, Merkle proofs against a single root vs Merkle proofs against
This would affect light nodes mostly in terms of processing fraud proofs. |
Define the process and parameters (number of samples) for doing Data Availability Sampling (DAS).
The text was updated successfully, but these errors were encountered: