Errors originating from infrastructure and their configurations are one of the major causes of system failures and system degradation, resulting in security vulnerabilities, application outages, and incorrect program executions. Investigating the root causes of such issues and remedies for them requires insight from different research perspectives, including systems, programming languages, software engineering, and verification. From these areas, approaches are emerging to manage the complexity of infrastructure and configuration, covering a breadth of forms, such as domain-specific languages, standalone verification tools, automated learning techniques, specification-based synthesis, security annotation extensions, and configuration optimizers.

The Dagstuhl Seminar on Resilient Software Configuration and Infrastructure Code Analysis brought together experts from different fields to explore new cross-disciplinary approaches to configuration management. The seminar facilitated collaboration between academia and industry and enabled synergies between different subareas of software systems. The seminar was a forum for cross-disciplinary discussions, bridged communities, and forged new conversations between academic and industrial perspectives. The shared knowledge built during the seminar is captured in this report, which we hope can act as a body of knowledge for researchers joining this newly forming community.

Overall, the seminar consisted of 3 tutorial talks, 16 presentations, and 5 group discussions. Emerging themes that were revealed during the seminar included a focus on Infrastructure as Code, the similarities and differences between configuration engineering and software engineering, the portability (or lack thereof) of program analysis techniques to configuration analysis, the design space of expressibility of configuration languages, and future challenges of analysis for safety, security, and auditing. In addition, we had a joint evening session with the parallel seminar "Agents on the Web" (Dagstuhl Seminar 23081), where each organizer presented an overview of their seminar. As a result, we started joint discussions where we investigated the use of formal methods, in particular synthesis, for establishing semantic relations between the data.

Key outcomes of this seminar were evident both in new short-term collaboration and connections, as well as the initiation of longer-term projects. For example, a collection of the participants have connected to host a workshop on configuration languages and analysis called CONFLANG 2023, which will be hosted at SPLASH 2023. Additionally, a vision paper outlining key future directions of the field is being drafted by participants of the event.

After two postponements due to COVID-19, this seminar was a pleasure to hold in person and a great success from both a community and a research perspective. We would like to thank the team of Schloss Dagstuhl for their hospitality and support as well as all the participants for their valuable contributions.

Creative Commons BY 4.0

Jürgen Cito, Ruzica Piskac, Mark Santolucito, and Andy Zaidman