Tests for validation of harvested data

[jeffreyliu]

What should the recipes return, and how should that output be tested? Should it be the responsibility of the recipe to validate, or will there be a separate service that validates the output? I'm leaning toward having a separate set of tests to validate output, but also have the recipe report some sort of schema or config which specifies what tests should be run and relevant parameters.

There are some existing tools for data validation, such as JHOVE, which we may be able to build off of.