During the HEP-CCE review and when implementing the RNTuple prototype in CMSSW, it became clear that the current REntry interface doesn't work for writing const objects (via raw pointers). This was confirmed by ATLAS in yesterday's RNTuple planning discussion. As REntry and RFieldBase::RValue heavily rely on shared ownership of the objects, a possibility would be to introduce a separate RConstEntry (name to be discussed) that only works for writing and acts as a sort of container for raw pointers only.