A persistent problem: managing pointers in Non-Volatile Memory

From the article:

The beauty of persistent memory is that we can use memory layouts for persistent data (with some considerations for volatile caches etc. in front of that memory, as we saw last week). Traditionally one of the major costs when moving data in and out of memory (be it to persistent media or over the network) is serialisation. With NVM we have both an opportunity to avoid serialisation overheads, and also an obligation since these overheads would be intolerable compared to the performance of NVM.
[...]
Traditional pointers address a memory location (often virtual of course). But the assertion here is that:

Persistent pointers, which implement references via identity and not ephemeral location, are vital for software to construct data structures in a world of persistent memory… Virtual addresses are insufficient as they are ephemeral and cannot encode identity-based data references for objects in a distributed, global address space without dramaticaly increasing pointer size or incurring coordination costs.