| 09:03 | <yhirano__> | annevk: hi |
| 09:05 | <annevk> | yhirano__: hey |
| 09:06 | <yhirano__> | Regarding https://github.com/whatwg/html/pull/3592, are you talking about https://html.spec.whatwg.org/multipage/browsing-the-web.html#traverse-the-history? or do you have other call sites in your mind? |
| 09:08 | <annevk> | yhirano__: the other one I'm unclear on is reloading |
| 09:08 | <annevk> | yhirano__: it says to use the Document object's URL, but it also takes into account other state such as whether the Document object was created as the result of a POST |
| 09:09 | <annevk> | yhirano__: which leads me to think it doesn't actually use a URL |
| 09:11 | <yhirano__> | annevk: https://html.spec.whatwg.org/#traverse-the-history gives a URL to the navigate algorithm. I can give a request with body and method set from the history entry instead, I think. As reloading reuses the previously used resource that will fix the reload case as well. Does it make sense? |
| 09:13 | <annevk> | yhirano__: I'm not sure; I suspect that's still wrong for browsers with a "fast" back implementation, where they keep the tree alive |
| 09:16 | <yhirano__> | sorry, what does the "fast" case mean? |
| 09:18 | <annevk> | yhirano__: Firefox doesn't always rebuild the node tree |
| 09:18 | <annevk> | yhirano__: so you go back and you end up with the same node instances as before |
| 09:18 | <yhirano__> | annevk: So are you talking about the case where the first step is skipped (i.e., Document is alive), right? |
| 09:19 | <annevk> | yhirano__: the HTML Standard supports this to some extent, though it's buggy |
| 09:19 | <annevk> | yhirano__: I'm not sure what you're referring to; none of these algorithms are particularly fresh for me |
| 09:20 | <annevk> | I wish I had managed to refactor these at some point so this wouldn't be such a problem, but I didn't succeed :/ |
| 09:28 | <yhirano__> | annevk: thank you, I will file a bug. I'm not sure I can / should fix this issue soon though. |
| 09:43 | <annevk> | Thanks, note that there might well be one already, there's loads of issues tagged with topic: navigate |
| 14:19 | <Domenic> | gsnedders: I might take over https://github.com/w3c/web-platform-tests/pull/5525 if you don't mind. Do you have any thoughts on whether the tests are still good according to your standards a year later? E.g. tests that take 5 seconds seem a bit suspect to me. |
| 14:20 | <annevk> | Domenic: the JS side of an interface callback is a non-function object, no? |
| 14:21 | <Domenic> | I am 95% sure that is the case, yes |
| 14:21 | <Domenic> | I see |
| 14:21 | <annevk> | Domenic: in that case the concern still applies unfortunately |
| 14:21 | <Domenic> | I remember us going through this in some incumbent global thing |
| 14:21 | <gsnedders> | Domenic: 5s is mostly just trying to avoid any chance of flakiness, esp. with debug, given it's essentially racing the network |
| 14:22 | <gsnedders> | Domenic: go ahead and take it over, though |
| 14:22 | <gsnedders> | Domenic: I only hadn't done anything because there's not a huge amount of interop and no spec |
| 14:22 | <annevk> | The bad thing with these complex things is that there's months that go by sometimes before a bit more progress is made and then having to remember it all is just... |
| 14:22 | <Domenic> | gsnedders: yeah, bz and I both seem to have gotten annoyed enough at this that we're going to make there be a clear spec. |
| 14:22 | <Domenic> | and tests |
| 14:23 | <Domenic> | annevk: So for https://heycam.github.io/webidl/#call-a-user-objects-operation we just decided to use the realm of the object, despite ES vs. reality concerns. |
| 14:23 | <Domenic> | It links to https://heycam.github.io/webidl/#dfn-associated-realm |
| 14:23 | <gsnedders> | Domenic: yeah, I saw |
| 14:23 | <annevk> | Domenic: hah |
| 16:42 | <annevk> | It seems the lower bar for an edge case is a 50-50 split across browsers |
| 16:42 | <annevk> | Not even too mad with such an outcome |