Code Speaks: Evaluating Japan's Local Government System Standardization

どうも〜おかむーです！ Today I'm taking a technical look at Japan's push to standardize local government core systems — "code speaks" style: read the policy, inspect the data and APIs, and suggest concrete engineering fixes.

• Local governments must adopt systems conforming to standards for 20 core tasks under the Digital Agency law
• There are public APIs (e-Gov, Tokyo Open Data) and dashboards (Japan Dashboard, e-Stat), but machine-readability and uniformity are inconsistent
• Practical fixes: API-first reference implementations, schema registry, automated data quality checks, and PDF-to-CSV pipelines

結論

Japan's legal framework (Digital Agency / Local Government Information Systems Standardization Act) sets a clear policy direction, but engineering reality is fragmented: some useful APIs and dashboards exist, yet many datasets remain PDFs or bespoke formats. 要するに、政策は正しいけど実装が追いついてないということです。エンジニア的に言うと、標準仕様とリファレンス実装があれば大半は解決するんですよね！

Report

Policy & current landscape

The Digital Agency's page and the e‑Gov law specify that local core business systems for ~20 services must be "standard-conformant". Practical assets already available:

• e‑Gov API catalog (administrative APIs)
• Tokyo Open Data API (PublicFacility etc.)
• Japan Dashboard & e‑Stat for national statistics

But these live in different islands. Many municipalities still publish PDFs or HTML tables — "これ見てくださいよ" — making automated reuse hard.

Machine-readability and API quality

Problems observed:

• Mixed formats: JSON/CSV APIs exist, but lots of PDFs remain (billing, meeting minutes, local budgets)
• No single API contract or schema registry across municipalities
• Authentication and rate-limits vary; discoverability is weak

Technical consequences: ETL overhead explodes, reproducibility suffers, downstream developers waste time parsing PDFs instead of building services.

Concrete technical proposals

API-first reference implementations

- Publish OpenAPI specs per standardized task; provide Dockerized reference server and client SDKs (Python/JS).

Schema & metadata registry

- Adopt DCAT + JSON Schema; publish machine-readable metadata at a central catalog (Japan Dashboard can host).

PDF → structured pipeline

- Use OCR + table extraction (Camelot/Tabula, Tesseract), then validate with JSON Schema; create a human-in-the-loop tool for accuracy.

Contract testing & CI

- Pact or schema validations in CI for vendors; semantic version APIs and require backward-compatible migrations.

Federated auth & API gateway

- OIDC for apps, API gateway for monitoring, quotas, and centralized docs.

Small code example

Here's a tiny Python sketch to fetch Tokyo's PublicFacility API and normalize into a DataFrame:

import requests
import pandas as pd
r = requests.get('https://portal.data.metro.tokyo.lg.jp/api/3/action/datastore_search?resource_id=PUBLICFACILITY&limit=100')
data = r.json()['result']['records']
df = pd.json_normalize(data)
df.to_csv('public_facilities.csv', index=False)

要するに、API一本でデータを引っ張ってくれば、CSV化・可視化は一瞬です。

まとめ

• Policy is strong: law requires standard-conformant systems for 20 core services
• Reality: fragmented formats, lingering PDFs, and uneven API maturity
• Fixes: OpenAPI refs, schema registry, PDF→structured pipelines, contract testing, federated auth

これらを進めれば、自治体の効率化と市民向けサービスの迅速化がぐっと進みます！

おかむーから一言

I've built and shipped products that integrate government data — the tech path is clear: standardize, publish, and iterate. Let's make government data actually usable, not just theoretically open!