Good-future-python charter¶

The template's founding idea — Good-future-python: generated projects ship with lawful, network-kind defaults enforced by code, not by documentation alone. Copyright, scraping politeness, and research ethics are constraints the toolchain checks (ruff / pytest / CI), so breaking them fails the build before it reaches review.

The three respects¶

Respect the source. Every web fetch goes through the polite fetcher layer (fetcher.py or the engine starter), which checks robots.txt before the first byte, sleeps between per-host requests, caches GET responses on disk, and identifies the project with a contactable User-Agent. Direct HTTP calls are banned project-wide by ruff's banned-api.

Respect the law. The project license stays attached (LICENSE + license-files); dependency licenses are audited by task license-check (pip-licenses --fail-on) against a copyleft policy the project license can absorb. AGPL dependencies (memorious) force the whole project AGPL. CAPTCHA-solving helpers are never generated — bypassing bot protection violates the site's terms. Personal data follows the same bar as DEIDENTIFICATION.md.

Respect the commons. Research software ships citation metadata (CITATION.cff) and SPDX annotations (REUSE.toml); datasets carry DUO use-conditions and CARE governance notes where they apply. Caches, harvests, and credentials are git-ignored — commit code, never data.

The legal surface¶

Two field rules name the law that the mechanical defaults sit inside. The scraping-law section (domain-scraping-law, on the scraping_effective gate) maps copyright and the EU database right (Directive 96/9/EC art. 7) against the contract question (ToS) and the unauthorized-access line (CFAA / 不正アクセス禁止法), with robots.txt as evidence of the operator's intent rather than an authorization — hiQ Labs v. LinkedIn (9th Cir. 2022) kept the contract claim alive while the CFAA reading narrowed. Its generated counterpart is LEGAL.md, the per-source pre-crawl checklist a scraping project ships. The contest-rules section (domain-contest-rules, on the oj_code judges) records that AI-use rules differ per judge and change (AtCoder bans generative AI during ABC/ARC/AGC except a whitelisted translation prompt), that the submission is the entrant's own responsibility, and that rating manipulation and multi-accounting are banned. Both are enforcement L0: they reach the reader through the generated AGENTS.md appendix, and the ethics-appendix predicate in tests/test_render_invariants.py holds them to the leaf classes that select them.

How it is enforced¶