Aaron Craelius

I think the one thing we might need here is a migration path for users of the previous encoding version. @EmilGeorgiev are you using the previous version in production? Even...

In the meantime should I submit a PR to disable the golang registration of bigdecimal?

Getting these warnings now: ``` 2021/12/15 11:12:07 Message cosmos.base.abci.v1beta1.StringEvent contains the reserved field name cosmos.base.abci.v1beta1.StringEvent.type which conflicts with protoreflect.Message interface implementation. This field will be suffixed with an underscore '_'....

Plus a bunch of errors: ``` cosmos/tx/signing/v1beta1/signing.pulsar.go:1154:42: v.ProtoReflect undefined (type *SignatureDescriptor_Data_Single has no field or method ProtoReflect) cosmos/tx/signing/v1beta1/signing.pulsar.go:1160:42: v.ProtoReflect undefined (type *SignatureDescriptor_Data_Multi has no field or method ProtoReflect) cosmos/tx/signing/v1beta1/signing.pulsar.go:1235:78: (*SignatureDescriptor_Data_Single)(nil).ProtoReflect...

Let's reopen this. I think it would be good to do the painful thing and not have to have the warning about fields getting renamed

If there is a type with a field method conflict, we should do codegen using a struct for the fast refletion wrapper instead of an alias to avoid the field...

So my proposed approach is to create a canonical marshaler in the cosmos-proto `runtime` that generated code will call on imported messages. Since the official marshaler is reflection based too,...

How would the fallback look like?

Basically anytime a message declares `extensions ... to ...` it would fallback? I also think it's fine to just fail.

Maybe an error is better for now since it's unintended usage?