SignedAvailabilityBitfields should be a typedef, not a newtype

[coriolinus]

polkadot/primitives/src/parachain.rs

Lines 727 to 728 in 90d5dbe

	#[derive(Encode, Decode, Clone, PartialEq, Eq, RuntimeDebug)]
	pub struct SignedAvailabilityBitfields(pub Vec<SignedAvailabilityBitfield>);

Rationale:

• We pass around binary encoded bitfields. A newtype may not encode to the same representation as its wrapped type; a typedef always will.
• The wrapped type is public and we do not implement any interesting traits on the newtype; we aren't gaining any advantage in type safety or ease-of-use.

I suspect this was simply an oversight, but we may as well fix it.

[bkchr]

We pass around binary encoded bitfields. A newtype may not encode to the same representation as its wrapped type; a typedef always will.

The encoding will always be the same, as SCALE does not encodes structs, just its fields.

The wrapped type is public and we do not implement any interesting traits on the newtype; we aren't gaining any advantage in type safety or ease-of-use.

You get type safety. That is mostly the reason why you choose a new type wrapper and it also often makes the code better to read.

[coriolinus]

You get type safety for cases like struct Meter(f32); struct Foot(f32);, which prevents you from adding meters and feet without an appropriate conversion. I submit that there is no Vec<SignedAvailabilityBitfield> for which it is inappropriate to interact with a SignedAvailabilityBitfields; the fact that the wrapped vector is public tends to support this assertion.

Here, it makes the code worse to read, because we need to add a sprinkling of .0 and .into() wherever we need to convert between one form and the other.