# Task-date interpretation measurements Edition: 25 October 2026. Original observations and fixture data are CC BY 4.0; original code is MIT. This is a synthetic parser comparison, not a transaction-system test or a legal deadline calculator. ## Inputs and configuration Twenty original strings cover date-only forms, explicit offsets, local times, spring gaps, autumn folds, valid and invalid calendar dates, sub-millisecond precision and alternative date notations. The full roster is `../examples/task-date-portability/fixtures.json`. The fixtures contain no people, transactions or legal dates. JavaScript uses `new Date(string)` in Node.js 24.19.0, with V8 13.6.233.17-node.51, ICU 78.3 and tzdata 2026b. `TZ=America/Los_Angeles` is set for that command only. Python uses CPython 3.12.14 and `datetime.fromisoformat`. If the result is naive, the runner attaches `ZoneInfo("America/Los_Angeles")` with `fold=0`. Explicit input offsets stay explicit. All Python values then convert to UTC and back to the named zone for the local-display field. The installed Python zone-data file hash, executable hashes and runtime details appear in `environment.json`; no hardware or local filesystem identifiers are published. ## Reading the fields Each row retains all fifteen blueprint attributes for both parsers. `parser_acceptance` is true only if parsing produced a value; it is not calendar validation. `native_parse_type` distinguishes Python naive/aware values from JavaScript's timestamp representation. `parsed_components` states whether it records Python's native parsed fields or JavaScript's post-parse local display. `explicit_offset` is a literal input suffix, or null. `assumed_zone` records interpretation, not a zone inferred for an actual transaction. `date_shift` compares the leading extended calendar date with the local-display date, where that input form exists. It is null for other forms. It is not automatically a defect. `precision` identifies the API's representation unit. `error` preserves an actual parse error or Invalid Date. Gap/fold fields are populated only for the fixtures designed to exercise those cases; null elsewhere means not applicable, not proof of absence. The Python gap row preserves both the attached wall-time representation and the local display after conversion through UTC. Attaching ZoneInfo does not reject the nonexistent local wall time. JavaScript resolves the gap during parsing. Both paths reach the same recorded UTC instant, but the intermediate operations differ. The autumn fixture uses Python fold=0; fold=1 is not tested as a separate naive policy. Explicit -07:00 and -08:00 inputs are included to expose the two instants. `roundtrip_effect` serializes each accepted value, reparses it with the same API, and records instant and original-text preservation. It does not claim that exporting from a vendor application preserves metadata. The joined index compares exact aware datetimes, retaining Python microseconds rather than rounding to JavaScript milliseconds. ## Evidence and reproduction Raw outputs are `node-results.json` and `python-results.json`; runtime/environment details are `environment.json`. Exact commands are in `../examples/task-date-portability/README.md`. Executable, fixture, runner and time-zone-data hashes distinguish the tested environment without redistributing binaries. The environment hash is SHA-256 of the compact, sorted-key JSON environment object using UTF-8. The raw outputs contain actual UTC timestamps. JSON is canonical. CSV text cells are formula-protected; JSON preserves original strings. Source documentation explains the APIs but does not substitute for the measured rows. The Python 3.12 web manual had advanced to 3.12.15 on the observation date, so the source ledger additionally pins the 3.12.14 documentation sources. No tests used a vendor account, real record, browser import or legal calendar. ## Source practice Each source's terms were read before its facts were compiled; sources whose terms prohibit compilation were left out; only short, independently worded facts and links are published; no descriptions, reviews, images or documentation are copied. ## Limits One build of each runtime, twenty deliberate strings and one named zone are covered. No pass rate, compliance score, platform recommendation or universal portability claim is made. Calendar-only dates and scheduled instants need distinct application schemas; this experiment does not decide any person's deadline. Repeat the fixtures after runtime, zone-data or import-policy changes and add original cases relevant to the intended application.