Embed Notice

@andre @alcinnz ... if you just put a sequence of edits into those chunks, all edits in one precede all edits in another, logically speaking. That's a lot less per-edit overhead, so less stuff to synchronize!

Unfortunately, it also loses enough precision to be potentially bothersome, so I devised a way to do better.

You can combine the logical, DAG-based ordering with a strictly monotonic wall clock per node. The nodes then attach a monotonic timestamp to each edit. It allows merging based...