In this section: Discovery and References
Research papers
Three published papers on distributed systems architecture by Enrico Piovesan. Each addresses a distinct problem: how services fragment, how portable execution boundaries work, and how governance applies when AI agents enter the execution path.
CSMA. Composable Software Model Architecture (June 2023)
CSMA addresses a specific architectural failure mode: distributed systems that fragment because service boundaries are defined at the deployment unit rather than at the behavioral contract. The paper introduces the composable service model (a structured separation between durable business logic and the runtime that hosts it) and proposes machine-readable contracts as the mechanism for making service composition inspectable and governable.
The three structural separations it defines (behavior from runtime, contract from implementation, placement from business logic) are the conceptual foundation this line of research builds on.
UMA. Universal Microservices Architecture (August 2024)
The paper that formalizes the model under its current name. The UMA white paper specifies the portable service as a WASM module with an active descriptor, defines the runtime layer's responsibilities (validation, adapter binding, trust enforcement, placement, execution evidence), and introduces behavioral equivalence as the measurable form of portability.
This paper is the architectural specification behind the book and the companion repository. The four structural properties it defines (explicit contract, portable binary, governed runtime, parity proof) remain the definitional core of UMA.
ECCA. Execution Context Coherence Architecture (August 2025)
ECCA addresses the governance question that emerges when AI agents become execution participants: how do trust, contract validation, and behavioral evidence apply when the caller is a reasoning model rather than a human-written service?
The paper introduces execution context coherence as a system-level property: the condition where heterogeneous participants (classical services, WASM modules, AI-assisted paths) produce verifiable, governed output regardless of which execution context initiated the call. It defines how Model Context Protocol fits into a governed execution model.