Hi A bit of context first: We have a zope application, using zope.schema to define our objects. The problem is, we have circular references, units contain references to persons and persons contain references to units. For schema validation, this is solved by subclassing zope.schema.Object and making it aware of reference objects and contained objects. So far so good. Now I'm working on z3c.schema2json (originally by Paul Wilson), allowing zope.schema defined objects to be converted to a dictionary structure. The problem I encounter here is that there is no awareness of reference or containment. This causes endless recursion going from person to unit back to person back to unit, . . A clean way to solve this, would be adding a marker interface IReference/ILazyLoad, . ., to the reference objects. If the serializer encounters the IReference interface, it can be skipped for serialization. The same concept could probably be used for the schema validation. To me the best place to add this marker interface is in zope.shema.interfaces, because it's actually working on zope.schema. I'm happy to fork zope.schema on github and make it working, but I'd like your feedback on this, before I start working on it. Pro's, con's? Scenarios I miss?