Privacy policy
Saved articles and helpful votes
We use visitor identifiers and cookies to save your choices. Press the same button again to remove a choice. When you are not signed in, your choices do not carry over to another device or browser.
Inquiries
Names, email addresses, messages and attachments received by email are used to respond to inquiries and address corrections. Do not send sensitive personal information or confidential third-party material.
Contact: kouyagino@proton.me
Details about stored data and cookies
Reaction records contain a hashed visitor identifier, article, reaction type and timestamp. The reaction database does not store names, email addresses or IP addresses. Account information supplied by the hosting platform is used to authenticate the site owner.
The functional cookie lasts up to 365 days. Deleting it does not automatically delete reaction records on the server and may prevent access to your previous choices. Reaction records currently have no automatic deletion deadline.
Public helpful totals and the owner's article-level totals do not display readers' personal details. Counts represent saved reactions, not verified unique people.
The hosting platform and external article-image providers may process access logs under their respective policies. Services linked from this site, including note, Substack and Amazon, apply their own policies.
Access measurement
We count page loads by Japan date, page/article and URL language. A random receipt for each page load prevents duplicate retries; it does not link different page loads. Receipts expire after24 hours and are removed on the next recording request.
Only if you choose “Allow cookies” do we use a random, site-specific cookie to estimate visiting and returning browsers. We store a hash of the identifier without linking it to names, email addresses or account IDs. This analytics database does not store IP addresses, User-Agent, search terms, URL queries/fragments or detailed referrer URLs. Only fixed referral categories are sent, such as Google search (inferred), Note, Substack, other external sites, internal navigation, or direct/unknown. Classification takes place in the browser; external hostnames, URLs and arbitrary query values are not sent or stored in analytics. An allowlisted utm_source link label takes precedence. Missing referrers may appear as direct/unknown; older records remain unclassified. This describes the preceding page referral, not a linked browsing session. The same opt-in identifier may appear under more than one category, so category browser counts must not be added together. We do not send analytics to advertising or external analytics services.
The measurement and preference cookies expire within90 days. Browser identifiers and per-day/page identifier records are used only within the90-day reporting window; expired records are removed on the next recording request. Page-view aggregates without browser identifiers are retained. When a DNT/GPC preference is received, visitor cookies are not used.
Use “Cookie settings” in the footer to select “Continue without cookies” at any time. This removes the visitor identifier cookie and stops future cookie-based measurement; past aggregate counts remain. You can read without accepting a visitor cookie. Estimates count browsers, not people. The owner report does not expose individual identifiers.
Authenticated owner visits, browsers marked for exclusion and declared verification requests are excluded from reader statistics. On owner browsers, a90-day exclusion cookie with a shared, non-identifying value keeps visits excluded after sign-out. It is separate from the visitor measurement cookie and does not link browsing history. Earlier mixed records are retained separately from reader totals.
To check measurement quality, we keep counts by public page, minute and fixed diagnostic category for14 days, removing expired counts on the next recording request. Categories distinguish owner or verification requests, declared automation and page loads whose origin is uncertain. Diagnostics store no IP address, raw User-Agent, cookie value or browser identifier and do not link page loads to the same visitor. Rapid browsing alone never excludes a reader view. Diagnostics start prospectively and do not change earlier page-view counts.