Clean up design for HTTP Accept header generation
Categories
(Core :: Networking: HTTP, enhancement)
Tracking
()
People
(Reporter: masterquestionable, Unassigned)
References
Details
Sending different Accept on context generally doesn't serve the intended purpose:
Sites really making use of which [1] are scarce (to almost none?).
And feature detection should be more reliable way for so.
[1] Mostly, making identical URL return different content based on Accept.
And the old behavior may cause unwanted side-channel.
This change removes "image.http.accept", making "network.http.accept" the sole control for Accept in all cases (besides Fetch/XHR) [2].
And reconfigured more reasonable default Accept.
Also cleans up some unused codes related.
[ [2]
Known 1 application:
https://github.com/orgs/community/discussions/70778
(GitHub's page loading)
; seemingly relied on the default `*/*`:
https://developer.mozilla.org/en-US/docs/Web/API/XMLHttpRequest/setRequestHeader
https://fetch.spec.whatwg.org/#ref-for-header-list-contains⑧
For which cause "network.http.accept.fetch" is added for debugging purpose. ]
|
Reporter | |
Comment 1•1 year ago
|
||
Actual change pending: I'm unfamiliar with committing code changes here.
Is someone more familiar with the Accept header able to provide some follow up?
|
||
Comment 3•1 year ago
|
||
(In reply to Master ? [:masterquestionable] from comment #0)
This change removes "image.http.accept", making "network.http.accept" the sole control for Accept in all cases (besides Fetch/XHR) [2].
And reconfigured more reasonable default Accept.
Also cleans up some unused codes related.
This phrasing and the unusual indentation and syntax makes me think the reporter is a bot. But looking into the substance of their arguments makes me think otherwise. 🤷
In any case, I don't think removing the prefs is a good idea. We'll need to send different headers for images vs page loads.
I think resolving bug 1711622 and bug 1865599 should get us closer to interop without a needing a different design.
|
|
Reporter | |
Comment 4•1 year ago
|
||
I use AI assisted content generation for most of my writing.
The format's design has solid basis. (to address various existing deficiencies)
Let's dig further then: via analyzing the current cases, and the theoretical boundary.
In brief, the current behavior creates undesirable side-channel. (and potentially bug-ridden due to over-complexity)
My aim is to address which without causing unjustifiable breakage.
Description
•