https://github.com/pk910/dynamic-ssz
Fastssz doesn't have as many features and isn't as actively maintained. Whereas, dynamic-ssz supports UnmarshalSSZReader which means we can switch to passing http.Response.Body, an io.Reader, directly to UnmarshalSSZReader, instead of having to hold the entire ssz byte buffer (310 MB on mainnet) in RAM while the struct is populated (reducing the peak ram requirement from 600+ MB by roughly half).
dynamic-ssz also may be more or less efficient at generating merkle proofs, so it should be benchmarked side by side with fastssz for the rp usecase.
https://github.com/pk910/dynamic-ssz
Fastssz doesn't have as many features and isn't as actively maintained. Whereas, dynamic-ssz supports
UnmarshalSSZReaderwhich means we can switch to passing http.Response.Body, an io.Reader, directly to UnmarshalSSZReader, instead of having to hold the entire ssz byte buffer (310 MB on mainnet) in RAM while the struct is populated (reducing the peak ram requirement from 600+ MB by roughly half).dynamic-ssz also may be more or less efficient at generating merkle proofs, so it should be benchmarked side by side with fastssz for the rp usecase.