Swift Networking Workgroup Meeting Notes
Meeting Details
- Date: September 28, 2026
Summary
The workgroup reviewed the new swift-networking-types repository, discussing the evolution process document and establishing contribution standards focused on performance and code size benchmarks. The main discussion centered on the HTTP client API strategy, where the group reached a consensus to prioritize building a native, top-to-bottom Swift networking stack over creating an abstraction layer for various platform-specific backends like URLSession. Next steps involve formalizing this strategy and the contribution process on the forums.
Notes
-
New Repository and Evolution Process
- The group discussed the new
swift-networking-typesrepository, which is now open. - An initial draft of the evolution process document was presented. The group debated whether this document should live in the Swift Evolution repository or on swift.org, similar to the SSWG incubation process.
- The process for changing the document was clarified: it would involve a PR, workgroup approval, and a final check by the ESG or core team.
- The group agreed that the formal evolution process is for stabilizing features, while initial work will be experimental.
- The group discussed the new
-
Experimental Features and Contribution
- There was a discussion on the duration and nature of the experimental phase. The consensus favored a longer experimental period to allow for real-world usage and feedback before APIs are stabilized.
- The group considered different mechanisms for experimental features, such as putting them in a separate experimental target instead of behind package traits, to avoid Xcode’s package resolution limitations.
- The group acknowledged the need for robust CI for any configuration, citing pain points from recent releases with experimental traits.
-
Contribution Standards and Benchmarking
- Jon proposed establishing clear contribution guidelines focused on performance, similar to WebKit’s historical standards, where PRs should not cause performance regressions.
- A key proposal was to establish a strong foundation of benchmarks from the beginning, tracking metrics like allocations, instructions, and especially code size impact, which is critical for embedded systems.
- The group discussed the challenges of accurately measuring code size due to compiler optimizations like inlining and specialization. The proposed approach is to use representative test targets and be willing to iterate or revert changes if they cause issues.
- The group agreed on the importance of defining high-level design principles, such as the trade-offs between performance, code complexity, and API elegance.
-
HTTP Client Strategy: Native vs. Abstraction
- The main topic was the layering and API strategy for the new HTTP client.
- Nakul framed the core question: should the group focus on abstracting over existing platform backends (like
URLSessionand AsyncHTTPClient) or prioritize building a fully native Swift networking stack? - A strong consensus emerged in favor of building a native Swift stack. Jon argued this is the only way to guarantee performance, control behavior, and avoid the “black box” limitations of libraries like
URLSession. - There was a suggestion that
URLSessionon Apple platforms should be considered out of scope for the workgroup, since the workgroup cannot direct its development. This reinforced the need to build a new stack from scratch. - The group acknowledged pragmatic challenges, such as the time and effort required for a full native implementation and special cases like the browser
fetchAPI for Wasm. - The possibility was raised that the new native Swift stack could become the underlying implementation for
URLSessionon Apple platforms, which would unify behavior and performance. - The immediate path forward is to design an abstract API that works with existing backends but is explicitly shaped to support a future high-performance native implementation, avoiding design bottlenecks.
Next Steps & Action Items
- Franz will set up the initial repository infrastructure for
swift-networking-types, including a mechanism for experimental features, before new code is merged. - Nakul and Jon will guide discussions on the forums and/or Slack to frame the HTTP client strategy discussion, outlining the consensus to build a native stack and proposing a high-level roadmap for community feedback.
- Eric will update the evolution process document, shifting focus to a practical
contributing.mdfile that outlines immediate contribution standards, including performance and benchmarking requirements. - Jon will attempt to locate the old WebKit design principles document to use as a reference for the workgroup’s own standards.