Limit meta tag parsing to one root, original tag #103
Conversation
This also adds support for speculative HTML parsing, whose fetches should respect the tag as well. w3ctag/design-reviews#702 (comment)
index.bs
Outdated
| ---------- | ||
| At [=speculative fetch=], after the sub-step for the meta element "viewport" step add a new sub-step with the following: | ||
| "A meta element whose name attribute is an ASCII case-insensitive match for "accept-ch" (only if this element is in the root Document and has not been modified by any script execution)." |
There was a problem hiding this comment.
What does "modified by script execution" mean? How would you expect someone to implement that? If the intention is to go find all the places in an implementation where a document can be modified by script execution, and have them flip some sort of boolean that you later consult... you need to find all those places in the spec, and do so!
index.bs
Outdated
| @@ -197,14 +201,13 @@ after step 6, add the following step: | |||
| Standard metadata names {#standard-metadata-names} | |||
| ------------ | |||
| For the section <a href="https://html.spec.whatwg.org/multipage/#standard-metadata-names">standard metadata names</a>, | |||
| add a subsection named `accept-ch` with the [=Accept-CH state|outlined explanation=]. | |||
| add a subsection named `accept-ch` with the [=Accept-CH state|outlined explanation=]. Note that this should be run exactly once, | |||
| and only on an element in the root Document which has not been modified by any script execution. | |||
There was a problem hiding this comment.
This does not seem interoperably implementable. When exactly should it be run? What if there are multiple elements with accept-ch? What if there are zero; how are we supposed to run this exactly once? How does this overlap with the speculative parsing section; if it's already run there, should it still be run here? (Plus the problem mentioned above about "modified by script execution" not being a defined term.)
|
I might be off base here, but are you trying to do something like the following? https://github.com/WICG/nav-speculation/blob/main/meta-processing.md Specifically the processing model at the bottom. |
|
I think the aim is similar? We want the browser to accept the meta tag at both the speculative and actual page rendering steps and apply the client hint delegations within to subsequent loads. However, only a single meta tag sent in the original HTML (pre-scripting) should be respected, and others should be ignored. |
|
I really encourage you to write a spec which mirrors the code you have implemented, so that others can follow that spec and also implement such code. I don't see how to go from what you have in this PR to anything like what you described in the comment you just posted. |
|
I'll be taking this up again as part of #108 |
This also adds support for speculative HTML parsing, whose fetches should respect the tag as well.
w3ctag/design-reviews#702 (comment)
Preview | Diff