Multiagent Foundations for Distributed Systems
Abstract: Early works and retrospectives by the researchers who founded the network protocols underlying current distributed systems indicate they were aware of the importance of capturing application meaning but didn't know how to handle it programmatically. Therefore, those researchers introduced simplifications in the protocols that violated their own principle of the end-to-end argument in systems design.
The thesis underlying this talk is the following. First, the above-mentioned simplifications, especially the reliance on reliable, ordered communication protocols such as TCP have run their course. Modern applications demand flexibility that can only be achieved through modeling application meaning, and many applications (such as those based on the Internet of Things) cannot pay TCP's overhead. Second, the multiagent systems community has developed alternative meaning-based approaches that can provide a new foundation for distributed computing at large. To support the thesis, I will demonstrate how modeling applications via information protocols enables realizing them over a transport as simple as UDP!
Biography: Amit Chopra is a Senior Lecturer at Lancaster University. He is interested in the software engineering of sociotechnical systems, including modeling, verification, and programming. Amit's work is founded upon representations of interaction between autonomous parties, via abstractions such as contracts, norms, accountability, and protocols. His work on contracts for distributed ledgers was adopted by a startup. Lately, Amit's efforts have primarily been directed toward interaction-based programming models that help developers create robust distributed applications easily, including on IoT and serverless platforms. Amit's recent work has been funded by the EPSRC.
