[thoughtlede]

I have implemented a JSON-like CRDT using this nice tutorial [1], which is a tree, but never an append-only DAG.

I'm curious how two edges meeting the same vertex will make it difficult to resolve the conflict automatically? Is it that two users will create the same vertex (concurrently) with slightly different attributes, and attribute-merging requires human-in-the-loop?

Of course if the DAG is not append-only, but supports, say, "moving" of edges/branches, it is a totally different problem. Martin has something to say about this [2].

[1] https://www.bartoszsypytkowski.com/operation-based-crdts-jso... [2] https://martin.kleppmann.com/papers/move-op.pdf

[preseinger]

> what will happen [if] we pick minimum values of the corresponding vector clock entries? The resulting vector clock will describe a point in (logical) time seen by both emitters. If we'll fold this minimum over all most recent timestamps received from all replicas, what we get is the timestamp that describes the point in time seen by everyone - so called stable timestamp. > > This timestamp serves us as a check point - since we know that all events prior to it have been acknowledged by all replicas.

this is not correct, as there is no reliable definition of "all replicas" -- nothing ensures that all replicas in the system are represented in the accounting for a given entity