Privacy

Traceary sets no cookies, runs no analytics, has no accounts and loads nothing from another host. What follows is the detail behind that sentence, including the parts that are outside our hands.

The short version

There is no sign in, no form that submits anywhere, no cookie, no analytics script, no advertising and no third party asset of any kind. The site is a set of static files. Reading it leaves nothing behind on our side that is tied to you.

What is stored in your browser

Two things, and each one only when you ask for it. Choosing light or dark at the foot of the page writes an entry called traceary-theme into your browser’s local storage, holding the word light or dark. Setting the theme back to system deletes it. It is read once when a page opens so the theme does not flash, and it is never sent anywhere.

Naming a product on the stack page writes the second, called traceary:stack, holding the list of products you named and nothing more. That list is the reason the page exists and it stays in this browser: it is written into the address of the page as well, so a stack can be sent to someone, and it is never sent to us. Clearing the list empties the entry. We hold no copy, and there is nothing here that could tell one reader’s list from another’s.

No cookie is set. The live site answers without a Set-Cookie header at all, which is something you can check yourself from a terminal rather than take on trust.

What the pages load

Everything a page needs comes from traceary.com. Fonts are served from here rather than fetched from a font host, so opening a page tells Google nothing. The collected release notes have their images stripped on the way in, so no page carries an image from anywhere, which also means no tracking pixel can ride in on someone else’s notes.

This is enforced rather than promised. Every response carries a content security policy that allows scripts, styles, fonts, images and connections from this origin only, and the policy is written in public/_headers in the repository. A third party asset would be blocked by your browser, not merely absent by intention.

The search boxes

The product finder on the home page works on a list that is already inside the page you downloaded. Nothing you type there leaves the browser.

The filter and the command palette fetch one file of rows, /browse.json, and they fetch the same file whatever anyone types. It says nothing about you.

Searching inside the notes is the one place where what you type has any effect on what is requested. The word index is split into 27 files by first letter, and keeping a word fetches the file for that letter, so the host serving this site can see that someone asked for a word beginning with o. It cannot see the word. Nothing is sent as a query, nothing is logged by us, and typing on its own fetches nothing: the file is requested only when a word is kept. Downloading all 27 files would hide even the letter and costs 653 kB, which is a worse trade for almost everyone, so the letter is the thing we tell you about instead.

What the host sees

The site is hosted on Cloudflare Pages, and delivering a page means a connection is made. Cloudflare processes that connection the way any host does, which includes your address, the page you asked for, the time and your browser’s user agent, and it uses it to serve the file and to absorb abuse. Cloudflare publishes its own account of how it handles that.

Traceary adds no measurement on top of it. There is no analytics product enabled, no visitor profile, no log we keep ourselves and no record anywhere that is tied to a reader.

One detail worth being exact about: Cloudflare attaches its own network error reporting header at the edge, so a browser that fails to reach the site can send a report about that failed connection to Cloudflare. It reports connection failures rather than reading, it goes to the host rather than to us, and it is not something Traceary receives or asked for.

Leaving the site

Release pages link back to their publisher, and the notes carry whatever links their authors wrote. Following one puts you on someone else’s site under their handling of data. Traceary sends a referrer policy that passes only the origin, so the publisher learns that you came from traceary.com and not which page you were reading.

If you write to us

Mail sent to hello@traceary.com is kept only as long as it takes to deal with, and it is not added to any list. There is no newsletter to be subscribed to. Ask and your message is deleted.

Since there is no account and no record tied to a reader, there is nothing held here to export or erase apart from a message you sent yourself.

If this changes

Adding anything that measures readers would make most of this page false, so this page is the first thing that gets rewritten and it is rewritten before the change ships. Traceary’s own changelog records changes of that size, and the terms of use and the imprint cover the rest.