Backlog triage: all 46 open issues and 2 PRs, one line each

[MichaelLeeHobbs]

Backlog triage — all 46 open issues + 2 PRs, one line each

Outside contributor, not a maintainer. The backlog looked large enough to be discouraging, so I worked through every open issue and PR and verified each claim against the shipped standard/*.json and the current published standard. No changes requested — this is offered as triage you can take or leave, and every line is checkable.

Headline: the published JSON is two years stale, for a fixable reason. standard/*.json is rev2024b (2024-04-18). The update-standard workflow is green — Apr/May/Jun/Jul 2026 all succeeded — but its final "Publish to master" step is commented out (c670936), so every month it regenerates the standard, uploads an artifact, tests it, and discards it. That single disabled step is the root cause of several issues below.

Second finding: 24 of 46 issues cannot be fixed in this repository — 15 File Editor, 9 website content. That is why the tracker looks unmanageable.

---

Fix first

#	Recommended action	Why
116	PR #130 is open for this	Workflow already succeeds; only the commit step is disabled. #130 re-enables it as an auto-opened PR, which answers the "unapproved commit" concern that caused it to be turned off. Tested end to end on a fork — the regenerated diff is +651,911 / −479,771 across 11 files, including 40 SOP Classes that have never been published.

Closes automatically once #116 lands

#	Recommended action	Why
126	Verify after regeneration, then close	LABELMAP appears 0× in shipped data — only because the data is rev2024b.
125	Verify after regeneration, then close	Reporter is right: a 2024b defect NEMA fixed in 2024c.
74	Regenerate, re-diff, then fix one real parser bug	621 rows shipped vs 655 in today's Table E.1-1; ~30 of the 35 postdate 2024b. Separately, the table yields a bogus row id: "ggggeeee-where-gggg-is-odd" from the "(gggg,eeee) where gggg is odd" line.

Close now — verified not a bug

#	Recommended action	Why
69	Close as fixed	Dnand occurs 0× in the current standard and 0× in shipped JSON; text reads "Dn and".
88	Close	All three Overlay modules are present (22/5/4 CIODs). Tags use 60xx placeholders, so searching "6000" finds nothing.
119	Close	(3006,0088) is marked RET (2022b) in PS3.6; the report links the 2015a edition.
123	Close	Every SQ has VM 1 — PS3.5 §7.4.3 says so explicitly. Multiplicity is Items, not VM.
129	Close	PS3.5 §7.4.4 gives 2C the same rules as Type 2, and §7.4.3 permits zero length when a value is unknown. The label is correct.
91	Close, label issue-with-standard	PS3.3 C.8.4.10.1.7 itself reads "measured in MegaBecquerels (Mbq)". Faithfully parsed; needs a CP to NEMA.
79	Close	pydicom usage question — group 0002 lives on ds.file_meta. Already answered in-thread.

File Editor — recommend discontinuing the feature (15)

15 issues, ~5 underlying defects, oldest January 2021, none triaged. Front end last deployed 2025-06-17, shipping React 16.13.1 (March 2020) — an unsupported major parsing untrusted binary input. Files are parsed in-browser and never transmitted; the concern is an unmaintained beta that looks supported.

#	Recommended action	Why
96	Close, and discontinue the File Editor	Reproduces (edit a value inside a sequence, then export). Fixing it still leaves the parse, encoding and validation clusters.
72	Close	Same defect as #96 — read vs. write failure on the same export path.
85	Close	No repro in the report; #96 has it.
124	Close	Same export defect as #96.
78	Close	Parse cluster. The editor renders Unable to parse <name>. for every failure, discarding the exception; parser is dcmjs, so some fixes belong upstream.
86	Close	Same parse signature as #78.
61	Close	Likely missing (0002,0000), the confirmed cause in #109.
57	Close	Secondary Capture files commonly ship minimal File Meta — same cause as #109.
84	Close	Valid: must honour Specific Character Set (0008,0005), not assume a default.
77	Close	Same as #84. The proposed unconditional UTF-8 decode would corrupt ISO_IR 100/144 and the ISO 2022 sets.
60	Close	Valid, and the worst of them — displays (3006,0010) under (3006,0080)'s label on a correctly-read file.
111	Close	Correct report, but unfixable as filed: no attribute-level conditions are published (module-level only, 315 of 3258 rows).
121	Close	Mixes the same conditions blocker with three plain structural violations.
120	Close	Needs a VR-level validation layer that does not exist.
109	Close	Resolved by the reporter — the file lacked (0002,0000).

Website content or UI — not fixable in this repository (9)

#	Recommended action	Why
87	Redirect to whoever owns the site	ISO_IR100 occurs 0× in this repo; no JSON file has any example-value field. Example values are authored on the site.
90	Redirect	Same. Values are legal Person Name syntax but poor examples.
73	Redirect	Same. "Houndsfield" is a site-side misspelling.
76	Redirect	Example value is site content; the unit it conflicts with is the standard's own — see #91.
42	Redirect	Website feature (deep links). Best value of the four site requests; a maintainer already agreed in-thread.
65	Redirect	Website filter feature. More useful once #49 lands, but the filter itself is site-side.
110	Redirect, with a workaround	Website export feature. module_to_attributes.json here already is the exportable table.
71	Redirect	CSS issue from 2022; worth re-testing before anything else.
58	Split	(a) CIOD sort order — website. (b) Transfer syntax listing — genuine gap here, duplicate of #122.

Parser work worth doing (6)

#	Recommended action	Why
49	Accept — highest-value enhancement	Enumerated values exist only as unstructured HTML inside description. Structuring them enables validators and would have made #126 machine-detectable.
122	Accept, retitle to "parse the UID registry (PS3.6 Table A-1)"	The repo models zero transfer syntaxes; sops.json is only SOP Class UID → CIOD. Supersedes half of #58.
66	Accept	Content Sequence recursion is truncated silently — (0040,A730) appears once, 0 paths contain it twice. Truncating is right; not signalling it isn't. Add a recursion marker.
48	Accept	Notes are not linked into tag descriptions; needs a table-row → note link.
63	Document the limitation, relabel enhancement	Print Management is a normalized IOD; the parser ingests only composite IODs (PS3.3 Annex A). Currently mislabelled bug.
43	Needs #49-style work first	All 9 macros are expanded into SR Document Content unconditionally, so Type 1 attributes are listed that may not apply.

Housekeeping (5)

#	Recommended action	Why
22	Close as done, or narrow-retitle	The example given — part-15 vs part-6 retirement consistency — is already implemented in verify_table_integrity().
31	Retitle to "Add TypedDicts for JSON row shapes"	The Python-version half is done (already 3.12); TypedDict remains unused, with one TODO at extract_attributes.py:14. Good first issue.
21	Fold into PR #100	There is no CI to run end-to-end tests against yet.
20	Finish the last bullet and close	5 of 6 checkboxes are complete; only "explain functional group macros" remains.
19	Re-verify against current, then close	Filed 2020 against NEMA's site; those defects have most likely been fixed upstream.

Open pull requests

PR	Recommended action	Why
#100	Rebase and merge	Still needed — there is no CI workflow, .travis.yml is still present, and README line 1 renders a dead Travis badge. Conflicts only on the requirements.txt hunk, which master has already superseded.
#108	Close (@dependabot close)	Obsolete — master already pins requests==2.32.3, newer than the 2.32.0 proposed.

---

I have posted the reasoning behind each recommendation as a comment on the individual issues, so nothing above has to be taken on trust. Every claim is reproducible against the current standard and the JSON on master; where I could not verify something I have said so rather than guessed.

Happy to open individual PRs for any of the parser items. And if it would help, I am willing to take limited permissions on this repository — triage only, enough to label and close issues — so the backlog can be cleared without costing you review time. Entirely your call, and no offence taken if the answer is no.