Using trait bound instead of a concrete session type bound

help
1

What are your thoughts on using struct SessionType<B : Bound> where Bound is implemented by only one struct ConcreteBound instead of just SessionType<ConcreteBound> directly?

E.g. instead of implemenating:

impl Deserialize for SessionType<ConcreteBound>  { 
 // ..
}

doing:

impl<V> Deserialize for SessionType<V> where V : Bound { 
 // ..
}

impl Bound for ConcreteBound {}

The primary motivation is that serde's derive macros do already support trait based bounds, eg. #[serde(bound(deserialize = "V: Bound")) while they do not support concrete type bound like #[serde(bound_type(deserialize = "ConcreteBound"))] I suggested here: derive `Deserialize` only for specific session types · Issue #2869 · serde-rs/serde · GitHub . So downstream consumers of SessionType could use serde's derive macro:

#[derive(Deserialize)]
#[serde(bound(deserialize = "V: Bound"))]
struct DownstreamType<V : Bound> {
   foo: SessionType<V>
  // ..
}

While with the concrete type, they are stuck manually implemenating Deserialize on a DownstreamType<ConcreteBound>.

Do you see any downsides or problems to using impl X for SessionType<B : Bound>?