In this talk I will discuss a simple, but powerful way to think about communication in distributed systems: whenever two (or more) processes interact, they dump all of their current local knowledge to each other. This full‑information view, standard in distributed computing, becomes subtle once we admit an undetermined environment that decides when communication happens and who participates. From the perspective of an individual process, a single interaction can then convey an arbitrarily large amount of information, so the traditional automata-theoretic approach breaks down.
To analyse such systems, we model them as infinitely repeated games with imperfect information. The central message will be that effective synthesis hinges on identifying the right support for game equivalence: although the process may face unboundedly branching information trees, many different information histories are strategically indistinguishable. By quotienting these histories via an equivalence that preserves winning strategies (and ω-regular objectives), we can replace the unmanageable information growth by a finite, strategy-faithful game abstraction.
The talk is based on joint work with Laurent Doyen and Thomas Soullard presented at FSTTCS 2025.
Short Bio: Dietmar Berwanger is a CNRS researcher at Laboratoire Méthodes Formelles in Paris-Saclay. He obtained his PhD in 2005 from RWTH Aachen University, and his work sits at the intersection of logic, automata theory, and infinite games, with applications to formal methods. In recent years, he focused on strategic coordination in games and distributed systems, in particular on distributed information. Strongly attached to India and its research community, he will be joining the Indo–French joint research laboratory ReLax in Chennai for the coming five years.
