Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

Review: PR #79

Overview

A pure mechanical rename of the store-and-forward connect-string key sf_max_bytes → sf_max_segment_bytes, including the public snake_case property on SenderOptions, the private field and its user-set flag, the two hand-maintained key lists, user-facing error messages, tests, and the README. No runtime behavior changes. Declared breaking (!) with no back-compat alias, matching the parallel Java client rename (java-questdb-client#70).

Correctness — verified complete ✅

• No leftovers. A case-insensitive sweep for sf_max_bytes / sfMaxBytes across all .cs, .md, .yaml, .json files (excluding build output) returns zero hits. Code, tests, benchmarks, examples, and docs are all consistent on the new name.
• All three key registries updated, which is where a rename like this typically breaks silently: the parse site (ParseLongWithDefault(nameof(sf_max_segment_bytes), ...), SenderOptions.cs:287), the BuildWebSocketOnlyKeys() literal list (SenderOptions.cs:928, drives both the string-ctor WS-only rejection and the ToString() skip), and QwpConnectStringKeys.IngressOnly (QwpConnectStringKeys.cs:67).
• ToString() round-trip holds. ToString() emits keys by reflected public property name and consults WebSocketOnlyKeys by prop.Name (SenderOptions.cs:2308-2311); since the property, the parse nameof, and the literal list all moved together, emit → re-parse stays consistent.
• Clean failure mode for un-migrated configs. A beta.1 user still passing sf_max_bytes hits RejectUnknownConnectStringKeys and gets Unknown connect-string key 'sf_max_bytes' as a ConfigError at construction (SenderOptions.cs:385-393) — fail-fast, not silently ignored. That makes the no-alias hard rename safe in practice.
• User-facing messages updated in both places they exist: the three validation errors in ValidateStoreAndForwardOptions and the oversized-frame hint in QwpSegmentRing.TryAppend (QwpSegmentRing.cs:328).

Test coverage ✅

• All test config strings and assertions migrated, including the two tests that pin the error-message content (Message.Contains("sf_max_segment_bytes")) — those would have caught a missed error-string rename.
• Test method names renamed to match (SfMaxSegmentBytes_NonPositive_Rejected, SfMaxTotalBytes_LessThanTwiceSfMaxSegmentBytes_Rejected). These don't intersect the CI's FQN suite filters, so no pipeline impact.
• Sf_AllKeysOnHttpScheme_RejectedIndividually still exercises the new key against the WS-only guard, so the registry update is itself under test.

Nits

• README table padding (README.md:315): the renamed cell kept the old column's trailing-space width, so the raw-markdown pipe alignment of that row is off relative to its neighbors. Renders fine on GitHub; trim if you touch the file again.

Verdict: approve. Complete, consistent, well-scoped rename with the right failure mode for stale configs; nothing blocking.