You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The algorithm 2 produces results that violate causal consistency when messages are duplicated and reordered by the network. This situation can occur in practice because the UDP protocol, which is used for the server-server communication, does not guarantee in-order delivery nor does it guarantee at-most once delivery.
The following steps are sufficient to reproduce the bug, with Client A, Client B and Client C running on different servers:
Req 1: Client A: PUT key, “NA”
Req 2: Client A: PUT key, “Request”
Req 3: Client B: GET key -> “Request”
Req 4: Client B: PUT key-effect, “Reply”
<replay packets sent by Req 1>
Req 5: Client C: GET key-effect -> “Reply”
Req 6: Client C: GET key -> “NA”
The steps above show that client C can see the effect event (“Reply”), produced by Client B, without seeing the cause event (“Request”), produced by Client A.
The text was updated successfully, but these errors were encountered:
The semantics in the paper models reordering of messages but not message duplication. Duplication can be modularly handled by transformers presented in Verdi paper. http://verdi.uwplse.org/verdi.pdf
The algorithm 2 produces results that violate causal consistency when messages are duplicated and reordered by the network. This situation can occur in practice because the UDP protocol, which is used for the server-server communication, does not guarantee in-order delivery nor does it guarantee at-most once delivery.
The following steps are sufficient to reproduce the bug, with Client A, Client B and Client C running on different servers:
The steps above show that client C can see the effect event (“Reply”), produced by Client B, without seeing the cause event (“Request”), produced by Client A.
The text was updated successfully, but these errors were encountered: