Beagle SCM

DIS-070: be/ JS get rejects scheme-less fs-path sources — trees.mkd example 5 jab get ../../jab/dog / jab get ../TKT-123 has no code path; wiki spells the form file:../proj only

trees example 5 (lines 199-210, post-2026-07-12 rewrite) pulls between worktrees with BARE relative paths: jab get ../../jab/dog (the main tree's sub as source) and jab get ../TKT-123 (a sibling wt as source). No code path: isRemoteSeed (get.js:261-267) admits a seed only with a scheme, an authority, or a .be-suffixed path; a bare relative path falls through to inRepoSeed (get.js:292-293) and parses as an in-repo path/branch — out-of-tree → NAVESCAPE/garbage. The wiki documents only the schemed spelling file:../proj?feat (GET.mkd:31). The Bro spells of the same example (get ///jab/dog, get //TKT-123/) are DIS-063's operand classifier; THIS ticket is the scheme-less fs-path SHELL form only.

Input

Context

• isRemoteSeed get.js:261-267 gates on scheme/authority/.be suffix; dispatchRow get.js:284-293 routes everything else to inRepoSeed.
• parseRemote's own localish test is WIDER than the gate: hasStorePath (get.js:95-96) counts ANY non-empty authority-less path — dead permissiveness behind isRemoteSeed; the two predicates must end up aligned whichever way the ruling goes.
• GET-038 resolveLocalSource (get.js:141-165) already resolves a worktree path down to its real store — the machinery works once routed; only the gate excludes bare paths.
• GET.mkd:31 file:../proj?feat is the only documented local-path source form; Verbs.mkd:27 routes a bare tracked-path token to restore, :29 any /-bearing token parses as a URI.
• DIS-062 treated the analogous POST fs-path (jab post ../../jab, the deleted example) as optional sugar, not spec; the rewrite added these GET twins, making the form spec-ish — hence its own ticket.

Goals

• RULING (gritzko): widen the gate (a scheme-less relative path that anchors a wt/store IS a local source) OR keep the gate and fix the post's shell twins to file: spellings (user edit) — pick one.
• If widened: jab get ../TKT-123 == jab get file:../TKT-123; precedence stays deterministic — an IN-tree path is an in-repo seed (restore/subtree), only an out-of-tree path holding a .be/wtlog anchor seeds from it, anything else NAVESCAPEs; no existence-probe ambiguity on tracked names.
• GET.mkd:31 + the Verbs bareword note updated either way; example 5's shell twins verified end-to-end.

Constraints

• Repro-first: two-wt fixture, get ../TKT-123 FFs the caller from the sibling; a tracked sub/dir arg must stay a restore (precedence leg).
• One routing story with DIS-063's explicit-operand classifier — do not grow a second local-source detector.

WIP

Design decisions

• (none yet)

TODOs

• obtain the ruling: widen isRemoteSeed vs file: spellings in the post
• if widened: repro fixture + the in-tree-path-wins precedence leg
• unify parseRemote.localish with isRemoteSeed (one predicate)
• GET.mkd:31, Verbs.mkd bareword note; re-verify the example 5 twins

Blockers and bummers

• Gates on the ruling; mechanical after that.

Outcome

• FILED 2026-07-13 (trees.mkd re-audit); nothing landed yet.