Skip to content

TASKS.md

Latest commit

 

History

History
78 lines (56 loc) · 3.95 KB

TASKS.md

File metadata and controls

78 lines (56 loc) · 3.95 KB
Raw

Decisions needed on the open questions

This tracks what's still needed from Luca on docs/open-questions.md, following the groups laid out in that page's "Grouping and priority" section. Group 1 there is already resolved in the spec text; everything below is open.

1. Validator story (group 2 — the real risk)

These are cases where every MUST can hold and the pixels/joins are still silently wrong. Decide: commit to a documented validator layer (formalizing what scripts/check_sp_ops_zarr.py already does as advisories), and which items become spec-level MUST language vs. stay validator-only advisories?

Ranked by blast radius: Q35 (raw tile footprint vs. per-image stage readout, up to 41 px), Q19 (library has no unique key; n:1 edge is really n:5 for controls), Q46 (labels vs. source image physical extent, silently off by 2x), Q30 / Q50 (stacked-but-unregistered image indistinguishable from registered; sp-ops:registration has no home in raw), Q49 (layout vs. tile-to-well transforms can disagree by 82 px). Lower blast radius, same family: Q37, Q32, Q27, Q22, Q51, Q15, Q5.

  • Decide yes/no on the validator layer.
  • For each item above, decide: promote to MUST language in the spec, or leave as a validator advisory.

2. Provenance: OME/Bio-Formats reuse vs. new sp-ops keys

Q6 (raw instrument provenance), Q10 (derived-element provenance), and the layout-provenance half of Q2 (measured vs. stage-readout layout) currently propose new sp-ops-private keys or tables. The alternative: reuse the OME data model (Instrument/Objective/Detector, per-image/per-plane metadata) and extend it only where genuinely missing, since RFC-8 already invites this.

  • Confirm this direction before Q6/Q10/Q2 get rewritten — it changes the shape of the fix compared to what's currently written in those entries.

3. Blank/background tile capability (carried over from the merged-spec PR, no Q id)

Converters currently drop blank/background fields of view used for illumination correction and background subtraction — no home for them.

  • Pick one: (a) a kind field on sp-ops:tile (field vs background), or (b) a sibling calibration collection.

4. Join/table model additions (group 4, additive — decide yes/no on each)

  • Q33: let on take a list of column pairs (peak-to-read join needs y, x, sigma).
  • Q28: add a coverage field to an edge, separate from cardinality.
  • Q11 / Q48: require plain label/parent-label columns instead of parsing an id out of a string, or membership riding on a numbering convention.
  • Q12: add a SHOULD for compartment membership recovered via spatial join at status: suggested.
  • Q23 / Q26 / Q34: a read-length field, an edge type from feature definitions to the tables they describe, and naming the "fusion" merged-table encoding that features.md doesn't currently name.

5. Editorial batch (group 5) — go-ahead to just apply these?

Q38, Q1 / Q9, Q39, Q45 / Q42, Q32 (delete the false sentence), and smaller fixes: Q41, Q44, Q47, Q24, Q21, Q20, Q40, Q36.

  • Confirm I should apply these directly (default: yes), or flag any to review individually first.

6. Documentation-only notes (group 6) — confirm no spec change needed

Q3, Q18, Q16, Q25 — each needs at most one documentation sentence, not a spec change.

  • Confirm (default: yes).

7. Housekeeping

open-questions.md's own stated convention: "an entry that is resolved moves into the page it fixes." Q7, Q8, Q14 are now resolved in the spec text but still sit in open-questions.md as historical entries (including inline <- Q8 / <- Q14 / <- Q7 annotations in the dataset store trees).

  • Decide: remove/migrate those three entries out of open-questions.md now, or leave them as a record until a broader resolved-entries cleanup.