Beagle SCM

DIS-074: bare post never implies the tracked URI — it always commits local, and refuses to advance cur's own branch

RULED 2026-07-15 (gritzko, law #2 of DIS-071): bare get and bare post imply the URI the worktree tracks. be/ JS post always commits into the LOCAL shard and advances a LOCAL ref (post.js:714-721); reaching anything else demands an explicit slot — a Host slot for a wire push (post.js:608-611, ssh/https only, a scheme-less //X excluded by BE-033) or a Query slot for a branch FF (post.js:618-622). So a worktree cloned from be://localhost?/project or from a sibling tree has no way to say "post back where I got this from": worktree:50-54 shows the flow (post "msg" commits and the tree goes ahead, then a second bare post fast-forwards the tracked main tree) and it has no code path. Worse, the bare advance is refused twice over: with no changes a bare post throws POSTNONE: no changes since base (post.js:666-673), and the nearest explicit spelling post ?<cur's own branch> throws POSTNONE: cur's own branch — nothing to advance (post.js:263-264) — the exact inverse of the bare post's meaning. Carries the bare-post half of the stale DIS-061; the EXPLICIT-operand half is DIS-062.

Input

Context

Post knows three targets, and none of them is "where I came from".

• post.js:608-611: only slots.host routes to pushRemote; a scheme-less //X is excluded (BE-033) and eaten as nav context by authorityRepo (loop.js:189-217).
• post.js:714-721: the commit path writes the local pack, then advanceRef on the LOCAL shard — the only advance a bare post can reach.
• post.js:618-622: the bare-advance arm requires a Query slot; a clean wt with nothing to commit reports POSTNONE (post.js:666-673) instead of advancing anything.
• post.js:263-264: post ?<cur's own branch> is REFUSED as "nothing to advance" — under worktree:50-54 that is precisely the motion a bare post performs.
• DIS-062 covers the EXPLICIT post ///jab/dog operand — same local-receive machinery, different trigger; share the implementation.

Goals

Bare post is the advance arm with its target defaulting to the track.

• Bare post with nothing to commit FFs the TRACKED URI — a branch ref, a local worktree, or a remote — to the worktree's base (worktree:50-54).
• Drop the own-branch refusal (post.js:263-264) for the implied/explicit-same-branch case — advancing cur's own tracked branch is the normal motion, not a no-op.
• Reuse DIS-062's local-receive path; FF-only, no forced updates, no non-FF advance.
• POST bare-row + Verbs gain the implied-target form; worktree's flow becomes reproducible end-to-end.

Constraints

The advance must stay a SEPARATE call from the commit — that is the whole point of the split.

• Commit and advance stay split (DIS-076): a bare post advances because it is an ADVANCE-shaped call, never as a commit side-effect. Do not resurrect commit-time FF.
• Depends on DIS-071 (the track URI must exist before it can be implied) and shares the receive path with DIS-062 — land after or with them.
• Needs a user ruling: does a bare post on a DIRTY tree commit AND advance? worktree:50-54 reads as two separate calls; the one-call reading is not assumed here.
• Repro-first: two local worktrees, FF-able divergence, seeded with jab get file:<path>.

WIP

Design decisions

• (pending the dirty-tree ruling in Constraints)

TODOs

• repro: clone A→B, commit in B, bare post in B → A fast-forwarded, B untouched otherwise
• repro: bare post on a clean wt tracking ?feature → the branch ref FFs to base (today: POSTNONE)
• resolve the bare post's target from the track URI; keep the explicit Query slot winning
• drop the own-branch POSTNONE refusal (post.js:263-264)
• ask the user: does a bare post on a DIRTY tree commit AND advance, or commit only?
• POST bare-row + Verbs target forms

Blockers and bummers

• Depends on DIS-071's track reader; the local-receive path is DIS-062's, which is itself blocked on BRO-024:52 (///→main-tree mapping).

Outcome

• FILED 2026-07-15 (track/base audit); nothing landed yet.