Consider implicit node / subtree adoption

[mikesamuel]

8.2 DOM Core says

These are the changes made to the features described in DOM Level 3 Core.

...

Nodes are implicitly adopted across document boundaries.

IIUC, before, taking a DOM subtree from <template>'s and same-origin <iframe>'s content documents and moving it into the DOM required an explicit document.importNode step.

That now happens auto-magically.

This means that insertChild and related methods can now take nodes from different Realms' documents.

Do we need to guard against DOM subtrees that were created using different policy sets?

May depend on how we address #42

[xtofian]

Somewhat related: DOMParser.parseFromString (https://w3c.github.io/DOM-Parsing/#the-domparser-interface) returns a Document. Should parseFromString be considered a HTML sink, and hence require TrustedHTML?

Conversely, DOMParser is used as the building block of HTML sanitizers (https://github.com/cure53/DOMPurify/blob/987a1e5569a7a4d4db97a145a51ff1ceafde9b89/src/purify.js#L380), in which context it's assumed that it's safe to pass untrusted HTML to it (as long as the resulting Document is not attached without further sanitization).

The DOM-Parsing spec has no "security considerations" section, so it's not quite clear whether the intent is that parsing untrusted HTML is safe.

[koto]

I've covered these document-creating functions in #50.

[koto]

Regarding node adoption, it's an interesting case:

https://gadgets.kotowicz.net/poc/xss/?xss=%3Cscript%3EsetTimeout(_=%3Edocument.body.appendChild(frames[0].document.body.children[0]),%201000)%3C/script%3E%3Ciframe%20srcdoc=%22%3Cimg%20src=https://www.google.com/images/branding/googlelogo/2x/googlelogo_color_272x92dp.png%20onload=%27alert(location.href)%27%3E%22%3E

In Chrome, it seems that the scripts execute with from their original realm even after implicit adoption. If FF, this is behaving as expected.

My take on this would be to consider this out of scope. Same origin documents can also just script you directly. It's not the exact same vector, as here the TT-ed document is actively initiating the potentially malicious action, but this feels like an intentional I'm-going-out-of-my-way bypass.

[koto]

Re: node adoption, we decided to put it out of scope for v1, and to add it as a security consideration:

https://wicg.github.io/trusted-types/dist/spec/##cross-document-vectors