“Defining Message Formats” is the title of a message posted on the Service Oriented Architecture mailing list [1] which attracted a lot of attention. The post summarizes the dilemma faced by solution architects when they have to define a service interface:
2. Create an XML-based message set based on the company's enterprise data model, which the company has invested a great amount of money into, and includes a very high percentage of all attributes needed in the business. In a nutshell, we've generated an XML Schema from ER Studio, and will tinker with that construct types that define the payloads for messages.
3. Use MISMO mainly for its entity definitions, but simplify the structure to improve usability. We benefit from the common vocabulary, but would ultimately define dozens of transactions over the coming years that MISMO has already defined.
I am arguing in favor of a variant of option 2 and provides details of how it can be implemented.
First, contrary to some architects recommend, I don’t think that an industry standard is a good option for internal integration. These standards tend to be bloated, ambiguous semantically or too generic. But most importantly, they tend to have one, grand view of the land when in fact different actors see entities In different ways. For example, in telecom, in the service layer a Service entity “uses” a Resource while in the resource layer, a Resource “hosts” Service(s). The TMF model models the first relationship but not the second.
Option 2 relies on the fact that the organization has an enterprise data-model. Sometimes there is a formally defined enterprise data-model with a significant effort allocated to maintenance and governance. But more often there is no formal enterprise data-model, instead the collection of all the data-models is the “de facto” enterprise data-model.
Data modelling can be done at several levels of abstraction: conceptual (or ontology), logical and physical. Another way to categorize the data model is by scope: often different sub-systems have different data-model. I call a “shared data-model” the intersection of the sub-system data-models. The “shared conceptual data-model” is in the realm of the enterprise architecture and business architecture. It translates in a shared logical data-model which is used by integration architects to create the service interfaces (XSD, Java or C# classes).
The “shared logical data-model” is the starting point for the definition of messages exposed by a service. But how to get from the data-model to the message format?
In the same message thread, Michael Poulin suggests [2] that each component can/should have a “personalized” view of the shared logical data-model. Based on the recommendation, the message format is the physical data-model (XSD) of the logical view used by the component. This is a sensible approach for message definition and often used in practice. The challenge is creating views of a logical OO model. While the concept is very common to the database area, it is not used at all in the OO domain.
The decision on how to define the API is project specific. In my experience I found that starting from a shared logical data model of the critical applications in the ecosystem to be a practical and extensible approach.
References:
1 Defining Message Formats. http://tech.groups.yahoo.com/group/service-orientated-architecture/message/14783.
2 Re: [service-orientated-architecture] Defining Message Formats. http://tech.groups.yahoo.com/group/service-orientated-architecture/message/14784.