Open jimmygchen opened 1 month ago
A few other clients have implemented this, and only serve the request if the node already have at least 50% of columns.
Fetching 50% / 64 columns from peers isn't necessary expensive - it's equivalent to performing 4 slots of peer sampling, which we do a lot more than this during head sync. However, this would allow HTTP APIs to requests the p2p network, which we want to avoid.
Probably best to implement this endpoint for nodes custodying >= 50% of columns for now.
Description
The
getBlobSidecars
beacon API needs a bit of work to make it functional for PeerDAS, as full blobs will no longer be stored in every node, and will be stored in the form of data columns.For supernodes and nodes with > 50% columns, this should be relatively straight forward, as the node would be able to reconstruct the blobs locally and serve them.
For regular full nodes, they would need to fetch at least 50% of columns, perform reconstruction before the blobs can be served.