Feature Proposal: Payload Compression

[o-nnerb]

Overview

Payload Compression means compressing the outgoing request body before it's sent over the wire (e.g. gzip), and setting Content-Encoding accordingly. This is distinct from the decompression support that already exists in the package (Session.disableDecompression() / Session.decompressionLimit(_:)), which only handles decompressing responses — a feature already provided natively by AsyncHTTPClient.

Scope

• Add a Session method (mirroring decompressionLimit(_:)) to enable request-body compression, e.g. Session.compression(_:), backed by NIOHTTPRequestCompressor from swift-nio-extras (already a transitive dependency, already used on the decompression side).
• Supported algorithms: gzip and deflate — the same two NIOHTTPRequestCompressor already implements, matching what response decompression already supports.
• Sets Content-Encoding automatically based on the chosen algorithm.
• Operates on the final serialized request body at the session/channel level, independent of which Payload source produced it (Data/JSON/String/File/Form/the upcoming PayloadEncoder from Feature Proposal: Payload Encoding #196) — compression (Content-Encoding, a byte transform) and payload encoding (Content-Type, value serialization) are orthogonal, per the Feature Proposal: Payload Encoding #196 scope decision.

Out of scope

• Algorithms beyond gzip/deflate (e.g. brotli, zstd) — not supported by NIOHTTPRequestCompressor today.

Milestone: 4.1.0.

[o-nnerb]

Body updated with a concrete scope: request-body compression (gzip/deflate via NIOHTTPRequestCompressor from swift-nio-extras, already a transitive dependency), distinct from the existing response-decompression support. Ready to be picked up as an implementation task.

[o-nnerb]

Implemented in #275 — Session.compression(_:) available in main. Closing as resolved.

[o-nnerb]

Superseded by #342 — the design here (Session.compression(_:) as a closed gzip/deflate enum, no decompression coverage, brotli/zstd excluded by NIOHTTPRequestCompressor) is being replaced by pluggable Compressor/Decompressor protocols that work uniformly across both executors, with brotli support and third-party algorithms. Closing this one as outdated; #342 is where the current design lives.