"Source browsing context" not well defined

[jeisinger]

In https://html.spec.whatwg.org/#source-browsing-context it says that every navigation has a source browsing context, which is the browsing context responsible for starting the navigation.

Is it possible to define this further. Especially, if a navigation is triggered from a script, what is the source browsing context? The responsible browsing context of the entered settings object?

[foolip]

Have you tested what the answer is in various browsers? Retrospec is the best spec :)

[jeisinger]

So far, I only tested blink, and the answer is that we don't have a consistent rule :)

[foolip]

Paging @bzbarsky, Savior of the Universe.

[bzbarsky]

So... I can't recall offhand what the source browsing context is used for (anything other than referrer?). I expect that Gecko tries to get referrer from "the thing that started the navigation", but that's a slightly vague concept at times.

Past that, the spec typically does define what the source browsing context is. The question is whether it matches browsers or not.

For example, if you set location.href you end up in https://html.spec.whatwg.org/multipage/browsers.html#location-object-navigate which says you use the incumbent settings object to get the source browsing context.

Or if you navigate by setting .src on an iframe, you land in https://html.spec.whatwg.org/multipage/embedded-content.html#process-the-iframe-attributes which ... doesn't actually define the source browsing context. That's a spec bug; it should define it.

For comparison, https://html.spec.whatwg.org/multipage/embedded-content.html#the-embed-element-setup-steps does specify (step 2 of the image/svg+xml steps): "Navigate the nested browsing context to the fetched resource, with replacement enabled, and with the embed element's node document's browsing context as the source browsing context." There's similar language in the processing of <object data=whatever>. So seems like iframe should do the same thing. I believe that's what UAs do here anyway; certainly Gecko does.

Not sure whether this helps or not. Basically, every spot that invokes the navigation algorithm needs to decide what this value should be....

[foolip]

@jeisinger, eager to attempt making some sense of this in a PR? :)

[annevk]

This is #1130 which arguably is a duplicate of this, except that it is a lot more clear.

[annevk]

@jeisinger are you okay with closing this in favor of that issue?

[jeisinger]

of course