SSWG - February 14, 2024

Server Workgroup Updates
1

Posting on behalf of @FranzBusch

Attendees

@adam-fowler, @FranzBusch, @ktoso, @joannis, @sebsto, @graskind, @tomerd, @finestructure, @tachyonics

2024 goals

  • Concurrency & Structured Concurrency adoptions @FranzBusch
    • Guide for structured concurrency
    • Overview of where we are standing with our libraries and the remaining work
  • Swift 6
  • Marketing
    • Think about two guides that either need to be overhauled or created from scratch
    • Build out integrated example project
    • Setup user group meetings @joannis & @sebsto
    • Show-case two server side projects/success stories
  • Tooling
    • Find a solution going forward for swiftly
  • HTTP server and middlewares @tachyonics @Joannis_Orlandos @adam-fowler @FranzBusch

Action items

  • Remaining packages that need to be pitched:
    • swift-asn1
    • swift-certificates
    • swift-otel
    • swift-nio-ssh, swift-nio-http2, etc. (Proposal to just collect them under the NIO umbrella)
    • swift-service-discovery
    • swift-async-dns-resolver
    • swift-cluster-membership
    • swift-kafka-client
  • Use the DM with the platform and ecosystem charter to share feedback

Discussion

  • @hborla and @Ben_Cohen joining us to talk about Swift 5.10 and 6
    • Swift 5.10
      • Achieves the full data isolation to ensure a data race language
    • Swift 6
      • Swift can now be used by setting the language version in the Package.swift
      • @FranzBusch :a: Sendable needs better clarification since we have potentially over annotated our libraries
      • @ktoso Documentation on swift.org is missing for Concurrency
      • @graskind existential any feature has been deferred out since there are potential future proposals that will ease the migration
      • @tkremenek come up with community wide migration plan
      • @Ben_Cohen Swift 6 is an opt-in language mode so nobody is going to be forced to adopt the language mode once they download the Swift 6 compiler
      • @FranzBusch :a: corelibs aren't annotated and they will cause problems
6 Likes
2

Can you elaborate on this line at all?

3

This is relating to the fact that corelibs-foundation and corelibs-dispatch on Linux are missing a ton of Sendable annotations compared to their Darwin counterparts. Adding them is often not trivial since the implementations actually need to be vetted to be thread safe before adding the conformance. As far as I know corelibs-xctest is properly annotated.

1 Like
4

Thanks for the clarification. You knew where I was headed. :slight_smile:

2 Likes