I think from the API surface we should have something like this

struct BlockHeader {
   chain_id: ULID,
    height: u32,
   ...
}

impl BlockHeader {
   
   pub fn to_bytes(&self) -> anyhow::Result<Vec<u8>> {
    ...
   }
   pub fn from_bytes(bytes: &[u8]) -> anyhow::Result<Self> {
    ...
   }
}

And the same for other types like Signatures, Block etc.

Also I think we should not split the definition from the structural perspective a common block from the genesis or final. They all blocks with the same structure. So I am proposing to add a methods like

impl Block {
  pub fn is_genesis(&self) -> bool {}
  
  pub fn is_final(&self) -> bool {}
}

And leave all the validation logic under the validate(&self) method, and not do any validation during decoding.

What do you think @cong-or ?

ok yes, I will encapsulate the logic in types as suggested. I assumed the ledger was going to be written as a contiguous block of memory hence thinking purely in bytes rather than types.

@cong-or

I assumed the ledger was going to be written as a contiguous block of memory hence thinking purely in bytes rather than types.

Yea, you are right, from the storage perspective it will be how you described.

BTW, I am going to move this immutable_ledger documentation to this repo, so we will have a direct access to the block.cddl file.
What do you think to have as a part of the from_bytes decoding logic to have a strict validation agains the cddl spec. The same as we did with the cip36 registration data ?
Or maybe just validate it in unit tests, but seems to me we need to have a strict check somewhere that this decoding logic conforms to the original design doc.

@cong-or just created a PR #72

Also forget to mention, from the testing perspective I think it make sense to make such decoding tests as property tests,
e.g.