We would benefit from using the newly introduced delegating value transformers:

• The complexity of the trait fields would be unlimited as no assumptions will be made about the sub-fields.
• The implementation of the transformer would be simplified because it would not have to take care of sub-field JCR I/O operations.

The only known drawback of the approach is that the aggregating transformer wouldn't be able to control the property names of the sub-nodes (handled with sub-fields/transformers). Then the sub-transformers should take care of correct naming and prefixes. For the simple case (one-property fields) we introduce the transformer which enforces 'value' as a property name.