Positive Dedalus programs tolerate non-causality

• We provide an initial investigation of the CRON conjecture by J.M. Hellerstein.
• This conjecture relates causality of message delivery to program monotonicity.
• We formalize the conjecture for the language Dedalus, that is inspired by Datalog.
• We show that positive programs remain correct when messages are sent into the past.

Declarative networking is a recent approach to programming distributed applications with languages inspired by Datalog. A recent conjecture posits that the delivery of messages should respect causality if and only if they are used in non-monotone derivations. We present our results about this conjecture in the context of Dedalus, a Datalog-variant for distributed programming. We show that both directions of the conjecture fail under a strong semantical interpretation. But on a more syntactical level, we show that positive Dedalus programs can tolerate non-causal messages, in the sense that they compute the correct answer even when messages can be sent into the past.