Related Website Sets

Let's consider situations where you have a series of related websites with different domain names, and you want to give site content access to third-party cookies when loaded in a third-party context inside other related sites (i.e., embedded in an <iframe>). Typical use cases are:

• App sites: A single application may be deployed over multiple sites, aiming to allow users to navigate between them seamlessly in a single session.
• Brand sites: A set of brand assets may be contained in a single site but then deployed over multiple domains, including session data relating to user preferences, customization, etc.

Third-party cookie access is commonly blocked by browser cookie-blocking policies. Still, you can work around it using the Storage Access API — see Using the Storage Access API for details.

Related website are a progressive enhancement mechanism that works alongside the Storage Access API. Supporting browsers grant third-party cookie access between websites in the same set without having to go through the usual user permission prompt workflow, once Document.requestStorageAccess() (or Document.requestStorageAccessFor()) is called. This results in a more user-friendly experience for users of sites in the set.

You should bear in mind that:

• The Chrome-only Document.requestStorageAccessFor() method — which allows top-level sites to request storage access on behalf of embedded origin content — is only supported on domains within a related website set. See Using the Storage Access API for an example.
• When Chrome first supported the standard Storage Access API (that is, the Document.hasStorageAccess() and Document.requestStorageAccess()methods), it required calling sites to be part of a related website set. This is no longer the case.

A related website set consists of one primary site and up to five associated sites.

A set is represented by a JSON structure. A hypothetical example is as follows:

{
  "sets": [
    {
      "contact": "email address or group alias if available",
      "primary": "https://primary1.com",
      "associatedSites": [
        "https://associateA.com",
        "https://associateB.com",
        "https://associateC.com"
      ],
      "serviceSites": ["https://servicesiteA.com"],
      "rationaleBySite": {
        "https://associateA.com": "Explanation of affiliation with primary site",
        "https://associateB.com": "Explanation of affiliation with primary site",
        "https://associateC.com": "Explanation of affiliation with primary site",
        "https://serviceSiteA.com": "Explanation of service functionality support"
      },
      "ccTLDs": {
        "https://associateA.com": [
          "https://associateA.ca",
          "https://associateA.co.uk"
        ],
        "https://associateB.com": [
          "https://associateB.ru",
          "https://associateB.co.kr"
        ],
        "https://primary1.com": ["https://primary1.co.uk"]
      }
    }
  ]
}

To use a set, its JSON must be added to the related_website_sets.JSON file available on the Related Website Sets GitHub repository, which Chrome then consumes to get the list of sets to apply RWS behavior to.

Each site in the set must also serve a .well-known file at /.well-known/related-website-set.json, which serves to verify the set structure and the relationship between the sites in the set.

The primary site's .well-known file must explicitly list out the full set structure. https://primary1.com in the above example would need a https://primary1.com/.well-known/related-website-set.json file similar to the following:

{
  "primary": "https://primary1.com",
  "associatedSites": [
    "https://associateA.com",
    "https://associateB.com",
    "https://associateC.com"
  ],
  "serviceSites": ["https://servicesiteA.com"],
  "rationaleBySite": {
    "https://associateA.com": "Explanation of affiliation with primary site",
    "https://associateB.com": "Explanation of affiliation with primary site",
    "https://associateC.com": "Explanation of affiliation with primary site",
    "https://serviceSiteA.com": "Explanation of service functionality support"
  },
  "ccTLDs": {
    "https://associateA.com": [
      "https://associateA.ca",
      "https://associateA.co.uk"
    ],
    "https://associateB.com": [
      "https://associateB.ru",
      "https://associateB.co.kr"
    ],
    "https://primary1.com": ["https://primary1.co.uk"]
  }
}

Each associate and service site needs to specify its primary site in a .well-known file. Each non-primary site in the above example (e.g. https://associateA.com) would need a /.well-known/related-website-set.json file like this:

{
  "primary": "https://primary1.com"
}

For full details of the process, JSON syntax, and other requirements for submitting sets, see the submission guidelines. Only domain administrators can create a set containing their sites.

Bear in mind that the .well-known files are also verified as part of the submission process, so they need to be put in place before the associated set is submitted.

Once a set is active:

• Requests from sites in the set (via Document.requestStorageAccess()) to access third-party cookies that belong to sites in the set are automatically granted, and no user permission step is required.
• Document.requestStorageAccessFor() calls can be made from top-level sites in the set to request third-party cookie access for other sites in the set.

RWS has been designed with security in mind. It would be disastrous if a bad actor site were able to claim to be part of a set and gain the privileges that entails. Lets consider a theoretical bad actor site, evilsite.example.com, and look at some examples of attacks it could try to make, all of which would fail:

• evilsite.example.com claims to be an associated site in another set: If a site claiming to be in a set (i.e. by listing a primary in a .well-known file) is not included in the set submission and/or primary's .well-known file, it won't get the benefits of being in the set.
• evilsite.example.com claims to be a primary site, and submits a set that includes some would-be victim sites: The submission process requires that .well-known files hosted by non-primary sites explicitly list out their primary. If this primary doesn't match the set submission (i.e. if the associated/service sites expect to have a different primary, or don't expect to be in a set at all), the submission will be rejected.
• site1.example.com and site2.example.com are intentionally in the same set, but site1.example.com gets hijacked by evilsite.example.com: The impact of a site hijacking attack within a set isn't any worse than it would usually be, once the other sites are updated accordingly:
  • The regular Storage Access API requires an active opt-in by the embedded site, so site2.example.com can stop calling document.requestStorageAccess() when it's embedded in site1.example.com, avoiding a CSRF attack.
  • Use of requestStorageAccessFor() requires CORS, so site2.example.com could choose not to respond with the appropriate CORS headers when network requests are coming from site1.example.com, thereby avoiding a CSRF attack.

• The Related Website Sets demo demonstrates how RWS is used.
• Also see Using the Storage Access API.

Two browser vendors oppose this specification. Known positions are as follows:

• Mozilla (Firefox): Negative
• Apple (Safari): Negative

• Storage Access API
• Related Website Sets on developer.chrome.com (2023)
• Related Website Sets: developer guide on developer.chrome.com (2023)