Land the optimistic local-edit-application layer on top of the M10 CRDT
foundation: frontend-side rope replica with local edit application,
daemon-authoritative broadcast, and bidirectional cursor reconciliation.
Keystrokes feel instantaneous because the local replica answers next-render
queries before the daemon round-trip completes, while the daemon remains
the single source of truth for conflict resolution and broadcast to remote
replicas.
Architecture beats:
- BufferMirror (src/buffer_mirror.rs) holds a per-frontend rope replica
with explicit cursor-staleness tracking. Every event that may move the
active cursor or swap the active buffer marks the mirror stale; the
next CursorByte from the daemon clears it.
- CrdtOpOrigin {OptimisticReplica(FrontendId), DaemonKey} routes broadcast.
OptimisticReplica skips re-application on the originating frontend
(already applied locally); DaemonKey broadcasts to all replicas including
source -- covers Lua-driven and generated-buffer edits that bypass the
optimistic path.
- Generated buffers (*help*, *workers*, *pmacs-instance*, *errors*) funnel
apply_edit output through queue_daemon_origin_crdt_op so post-attach
CRDT upgrades don't drop their edits.
- forbid(unsafe_code) preserved throughout; loro 1.12 added as the CRDT
engine.
Audit posture: M10.10 shipped through six post-audit review rounds with
twenty-eight cumulative findings, most categorized as "incomplete
application of a prior round's mechanism." The audit doc records
grep-driven exhaustiveness as the standing countermeasure.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
M10.1 produced an API survey doc as its working artifact (the audit-trail for
the loro-vs-yrs benchmark evidence). The existing patterns covered AUDIT.md /
SHIP-GATE.md / PREREQUISITES.md but not the survey shape; future research /
decision tasks producing similar empirical-evaluation docs are also covered
by this generalization.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Land the Model Context Protocol (MCP) integration as a transport binding,
not a built-in feature. Six Lua functions plus userdata methods expose
the substance of three MCP feature areas (resources, tools, prompts), a
notification dispatcher, and a non-trivial AI-assistance example package
that meets the architectural ship gate (spec/pmacs-spec.tex:1572): zero
direct calls into the Rust core, zero special-cased MCP handling outside
the public API, source under 2000 lines of Lua.
The M9.5 -> M9.6 -> M9.7 -> M9.8 layered composition validates the claim
"AI is a transport binding, not a feature" -- pmacs-mcp-ai composes with
pmacs-mcp-prompts.render and inherits notification handling transitively
through M9.7's package, demonstrating that the AI domain is a layer
above MCP, not a thread woven through the core.
Subtask shape:
M9.1 stdio transport + initialize handshake + restart policy
M9.2 resources with in-flight + settled cache and per-uri invalidation
M9.3 tools with isError-vs-JSON-RPC-error semantics + cancellation
M9.4 prompts with required-argument validation
M9.5 notification dispatcher (on_notification, off_notification)
M9.6 tools-as-commands fixture package + 12 audit findings disposed
M9.7 prompts-as-result-buffers fixture package + tree-sitter-md grammar
M9.8 AI-assistance fixture package (363+ LoC; 17/17 acceptance tests)
M9.9 formal package audit -- PASS on all three criteria
M9.10 release: TRANSITION-M9.md + MCP-for-package-authors guide
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Levi Neuwirth