Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
43 changes: 34 additions & 9 deletions Documentation/ReleaseRoadmap.md
Original file line number Diff line number Diff line change
@@ -1,12 +1,35 @@
# Execution update — 11 November 2026
# Execution update — 12 November 2026

1.9.0 is published at `f2a0d12da7028edfbcb0cb307f19a921a3a12d03`. It delivered the
bounded structural wave, including the public read-only store, and lowered the Apple
deployment minimums to macOS 15 and iOS 18 (see the
[1.9 qualification record](Qualification/Release1.9.md)). 1.8.0 delivered located
root-array streaming. The next proposed wave is 1.10.0; it is proposed scope only.
Persistence and new query grammars are not included in any release so far.
The proposal and historical inventories below remain source-bound.
root-array streaming.

On 12 November 2026 the maintainer approved the revised train below, from 1.9.1
to 2.0.0. It replaces the earlier 1.10 persistence wave:

- **Swift-JSON adds no database or persistent store.** Domain stores (Swift-MCP-Server
tasks, Content Collections) live in their own repositories on the shared
[Swift-SQLite](https://github.com/ThorFuchs/Swift-SQLite) base. Swift-SQLite does
not depend on Swift-JSON; the domain store connects the two. Swift-JSON supplies
the exact bytes those stores need: RFC 8785 canonical JSON and a versioned envelope.
- Each release still needs its own design admission for new public API, finite
profiles, fixtures and complete CI on the exact source. Approval of the train is
not approval of a particular declaration.
- The developer kit (Claude Code plugin and skill) is updated in the same PR train
as every minor release, so it targets the latest published version.

| Release | Approved direction |
| --- | --- |
| **1.9.1 — hygiene** | Already on main: the `platform-floor` CI lane, the CLI source identity without DocC catalogs, branch housekeeping and developer kit 0.2.0. Tag/publication remains a separate decision. |
| **1.10.0 — exact interchange** | RFC 8785 JSON Canonicalization Scheme (canonical bytes for hashing, signing and content addressing); a versioned envelope built on it; structural diff producing an RFC 6902 patch; RFC 7464 JSON text sequences in JSONStreaming. |
| **1.11.0 — queries and diagnostics** | RFC 9535 JSONPath with finite budgets, kept separate from Transform semantics; detailed and basic JSON Schema output formats. |
| **1.12.0 — API consolidation** | Review every public surface before 2.0: naming, error and limit consistency, deprecations with replacements and migration notes. No break in a minor release. |
| **2.0.0 — compatibility contract** | Unchanged from the gate below: one documented compatibility, resource and platform contract for the accepted family, with migration guidance for every recorded break. |

The historical proposal and inventories below remain source-bound. Where they
conflict with this update, this update governs the order and scope of 1.10 onward.

# Release roadmap: 1.7 through 2.0

Expand Down Expand Up @@ -70,11 +93,13 @@ A major version is a deliberate compatibility commitment: record and justify any
actual break and provide migration guidance; do not manufacture a break to reach
2.0. Dates follow exit gates, not a promised calendar.

The 1.10 entry and later are **proposed scope**, not API approval or permission to start
implementation. All planned packages at 2.0 means the explicitly accepted release
portfolio. Lexical/semantic retrieval, embeddings, persistent indexes, weighted
graph algorithms, set algebra, matrix arithmetic, JSONPath/another query grammar,
new backends and text-sequence framing have no automatic place in this promise.
The 1.10 entry in this historical table is superseded by the execution update above.
All planned packages at 2.0 means the explicitly accepted release portfolio.
JSONPath (1.11) and text-sequence framing (1.10) are now admitted as roadmap
direction, each still subject to its own design admission. Lexical/semantic
retrieval, embeddings, persistent indexes, weighted graph algorithms, set algebra,
matrix arithmetic, other query grammars and new backends have no automatic place
in this promise.
If a later scope review admits one, give it its own product-boundary decision,
finite profiles, fixtures and release gate. Do not rename speculative designs
into committed packages.
Expand Down
Loading