Clicking a link should preconnect to destination as first party, not third-party. Otherwise, the preconnect is wasted when session isolation based on registrable domain is enabled.
Created attachment 443200 [details] Patch
Comment on attachment 443200 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=443200&action=review Does this make it so that script can use a.click() to preconnect in the context of another page? > Source/WebCore/html/HTMLAnchorElement.cpp:544 > + if (completedURL.protocolIsInHTTPFamily() && ((frame->isMainFrame() && isSelfTargetFrameName(effectiveTarget)) || isBlankTargetFrameName(effectiveTarget))) { I assume this makes it so that we don't preconnect for a link that is supposed to open in an iframe, right?
(In reply to Alex Christensen from comment #2) > Comment on attachment 443200 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=443200&action=review > > Does this make it so that script can use a.click() to preconnect in the > context of another page? I think it does. But since this will cause a top-level navigation, we would anyway connect to that server as first party, no? > > > Source/WebCore/html/HTMLAnchorElement.cpp:544 > > + if (completedURL.protocolIsInHTTPFamily() && ((frame->isMainFrame() && isSelfTargetFrameName(effectiveTarget)) || isBlankTargetFrameName(effectiveTarget))) { > > I assume this makes it so that we don't preconnect for a link that is > supposed to open in an iframe, right? Yes, since we are now preconnecting as first party and since we want session partitioning, it only makes sense to preconnect if clicking the link is going to cause a top frame navigation.
Committed r285224 (243848@main): <https://commits.webkit.org/243848@main> All reviewed patches have been landed. Closing bug and clearing flags on attachment 443200 [details].
<rdar://problem/84988999>