...
As an example, Ranger needs to know how a particular entity is classified so that the classification can be used within a policy (rule). Atlas has a complex graph oriented model, within which classifications can be multi level - for example a column may be classified as "employee_salary" whilst employee_salary may be SPI. Ranger however just needs to know that employee_salary is SPI, not how we got there. So we convert this complex model into something much more operationally focussed and deliver that over the API. The implementation will follow this graph, and build up a list of all tags that are appropriate to use. Note that in the case of Ranger it is actually the tagsync process that will call the GAF for this classification information, .
Ranger can do this today, but via a large number of individual requests to retrieve types & entities. Rather than these lower level queries to Atlas, in GAF we can offer result sets to make queries more efficient, and more appropriate notifications.
...