The published language of the open-host service (OHS) can be seen as an integration-oriented model to help simplifying the public interface of the OHS consumers. The published language is integration-optimized and exposes data in a more convenient model for consumers, specifically designed for integration needs.
When we are promoting the idea that each bounded context has its own internal (canonical data) model, the OHS in fact decouples the bounded context's internal model from the model used for integration with other bounded contexts. So the bounded context internal model can evolve without impacting the consumers of the OHS.
Say we want to design the integration model of the OHS or even multiple OHS's. Don't we have here the concept of the old school canonical data model we used for integration in the ESB / SOA era? In fact one can say that designing the integration models for the public OHS even contributes to the concept of the interchange context: a separate bounded context mainly in charge of transforming models for more convenient consumption by other components.
So question is: are we not going back to the 'old school' canonical data model from SOA ESB era with the concept of interchange context as a separate bounded context in charge of transforming models for more convenient consumption by other components? If not, what is the difference?