Reserved paths - Platform reference

Platform files

These three files are answered on the root of every host, before the deployment is read. The last column says what decides the answer.

Path	Answer	Answered by	Notes
`/favicon.ico`	The RTEdge.net logo as SVG. A request referred by a `.js`, `.css` or `.html` file is redirected to an emoji icon, and a request referred by a worker URL (`/:`) to a gear icon.	Platform	The exact path only. A host-root route never serves a `favicon.ico` export, but path routes are unaffected, so `/docs/favicon.ico` is yours. Give pages an icon with `<link rel=icon href=…>`.
`/apple-touch-icon*`	A 307 redirect to `https://rtcode.io/.svg`, the RTCode.io mark. The query is dropped.	Platform	Every path that starts with `/apple-touch-icon`. Deployments cannot serve files with these names; point `<link rel=apple-touch-icon href=…>` at a file with another name.
`/robots.txt`	A fixed text policy: a notice that automated collection is prohibited except as the content-usage directives allow, then `User-Agent: *`, `Content-Usage: tdm=n, search=y, inference=y`, `Allow: /` and `Crawl-delay: 10`.	Platform	The exact path only. A host-root `robots.txt` export is never served. A path route's own `/docs/robots.txt` is not intercepted, but crawlers read only the file at the host root.

The platform matches these patterns on the raw URL, on every URL of every platform host, whether or not anything is deployed there. Most of them match anywhere in the path, not only at the end, so they also claim worker paths such as /:api::save. They are answered before your deployment is read, so a worker cannot override them.

Pattern	Where it matches	Answer	Answered by
`::s`…	Anywhere in the path. The letters after the last `::s` choose the action.	The sign-in actions `::si`, `::sa`, `::su` and `::so`, described in Adding sign-in. Letters that name no action return a 501 page, "Invalid affix". The related forms `::sit`, `::sud` and `::sh` are Experimental.	Platform
`::i` `::e` `::c`	A path that ends in one of them, optionally followed by word characters, such as `::edit` or `::code`.	A 303 redirect to the source view, `?`. Experimental	Platform
`::v`	A path that ends in `::v`, optionally followed by word characters, such as `::versions`.	A 303 redirect to `??`, the version viewer. Experimental	Platform
`::u` `::t`	A path that ends exactly in `::u` or `::t`.	A 303 redirect to usage analytics or tail logs. See Usage and tail logs. Experimental	Platform
Other `::i` `::e` `::c` `::u` `::t` `::v` paths	Any other path that contains one of these after `::` and does not end in `::clear`.	An empty 404, with no body.	Platform
`/::edge`	Anywhere in the path, with a slash before `::edge`.	Diagnostics about how the request was served, as JSON. Experimental	Platform
`::clear`	At the end of the path.	Clears this site's data in the visitor's browser. Experimental	Platform
`::qr`	At the end of the path.	A QR code as SVG. Experimental	Platform
`?::/h:`	The raw query string contains `::/h:`.	The response of the URL with its headers changed. Experimental	Platform

Warning: Avoid :: in route paths, file names and worker paths. A path such as /a::cool is read as an edit shortcut and redirects to /a?, /::sfoo returns a 404, and /:api::save opens a sign-in page and never reaches your worker.

Reserved on every route

After the platform has had its turn, the edge router answers these paths and queries for every deployment. Request routing shows where they sit in the order.

Pattern	Answer	Answered by
`/::owner` `/::kv` `/::meta`	Whether you own the route, the KV viewer, and the deployment digest. The text after the first `:` must be exactly `:owner`, `:kv` or `:meta`. Experimental	Router
A trailing `*`	The index of the deployments you can edit under that host or path. It needs sign-in.	Router
A URL ending in `??`	A 307 redirect to the version viewer, even for a route that was never deployed.	Router
`?` and a URL ending in `/?`	The source view, which opens the deployment's source in RTCode.io. The forms `?io` and `?sync` Experimental do the same.	Router
`?m=uo`, `?ogo`, `?<ext>` and the social preview parameters	An ownership check, the generated social tags alone, a content-type override for the page, and per-link preview overrides. See Query parameters. Experimental	Router
Any other `:`	Everything after the first `:` becomes the worker's pathname. A route without a worker answers 404, "Worker does not exist."	Your worker

The URL suffixes reference describes each of them.

Canonical redirects

Before routing, the platform rewrites some URLs. All of these are 307 redirects, and they apply to every method.

Request	Redirected to	Notes
`/docs`	`/docs/`	A path with no trailing slash that contains no `.`, `*` or `:`. The query is kept. The new URL belongs to the route `name/docs`, not to the route `name`.
`/docs?`	`/docs/?`	For document and iframe navigations. This keeps the bare `?`, so `name.rt.ht/docs?` opens the source view.
`/docs:`	`/docs/:`	A path that ends in a single `:`, so that it reaches the worker of the route `name/docs`.
`/docs/*`	`/docs*`	A path that ends in `/`, other than `/` itself.

Paths are also normalized before routing, so //nav.js, /x/../nav.js and /%6eav.js all serve /nav.js. File names are case-sensitive: /NAV.JS answers 404, "File does not exist."

Names to avoid

The platform refuses requests whose path contains any of the strings below, ignoring case, before they reach your deployment. They count anywhere in the path, so they also apply to route paths, export names and worker paths. A route named name/php-notes, for example, is unreachable.

String	Names it rules out
`php`	`index.php`, `info.php`, `/php-notes/`
`wordpress`	`/wordpress/`, `/my-wordpress-site/`
`/wp-`	`/wp-admin/`, `/wp-content/`, `/wp-notes.html`
`/cgi-bin/`	`/cgi-bin/test`
`/.well-known`	`/.well-known/anything`
`/.git`	`/.git/config`, `/.gitignore`
`/.env`	`/.env`, `/.env.local`
`/.aws`	`/.aws/credentials`

Note: Check a name against this list before you build a site around it. A refused request never reaches the router, so none of the pages in Error pages and status codes describes it.

Export names

The name in o: becomes the last segment of a URL, so a few more rules apply to the files you export. Stick to letters, digits, _ and - plus one ., or to leading-dot names such as .css and .js.

Rule	Why
The name contains a `.`	A path with no `.`, `*` or `:` and no trailing slash is redirected to its slash form (see Canonical redirects), so a dotless name can never be reached.
The name has no `/`	A file exists only directly under its route. `/route/sub/x.js` belongs to the route `route/sub`.
No `:`, `?` or `#`, and no trailing `*`	A `:` sends the path to the worker, `?` and `#` start the query and the fragment, and a trailing `*` asks for the index.
No `::` suffix pattern	See Platform suffixes.
`favicon.ico`, `robots.txt`, `apple-touch-icon*` on a host-root route	The platform answers these paths first (see Platform files). A path route can export them under its own path.
The name matches in case	File names are case-sensitive, and the query is ignored: `x.js?v=2` serves `x.js`.
No spaces or non-ASCII characters	The URL is compared with the name in its percent-encoded form, so names such as `my file.txt` or `café.txt` may not match.

See Exporting files with o: for how exports are served.

Route names

See Routes for the recommended grammar, normalization and examples.

A path segment of a route cannot contain a .. When you type a route whose last segment contains one, it is taken for a file name and dropped.
A name.rt.ht subdomain has at least 4 characters, does not start with _ and is not a reserved word. See Domains and subdomains.
The apex rt.ht is not a deployment. It serves short links.

In the playground

The playground output does not apply these reservations, except ::su, which it answers itself. There, /favicon.ico, /robots.txt and :: paths other than ::su reach your worker. On RTEdge.net they never do, so test such paths on a deployment. Playground and edge differences lists the rest.