| 07:56 | <Noam Rosenthal> | Welcome bashi! a question to you and Yoav Weiss about early hints: do they survive same-origin redirects? |
| 07:58 | <bashi> | Noam Rosenthal: Our implementation is that they don't survive redirects including same-origin. |
| 08:01 | <Noam Rosenthal> | I personally think this could be useful - even if it's not implemented yet. e.g. you submit a form which is going to do some heavy processing and then redirect you to somewhere - might as well start sending the 103s beforehand |
| 08:03 | <bashi> | I think changing the behavior isn't difficult. I think we can update our implementation if that's preferable and there is consensus. |
| 08:06 | <Noam Rosenthal> | bashi: OK, since the reserved client survives same-origin redirects, for now we'll leave it in the draft PR and we can discuss the specifics on Github |
| 08:37 | <annevk> | I guess the question is whether the preloads are preserved? |
| 08:39 | <Noam Rosenthal> | annevk: exactly, if you started preloading something and then redirected to somewhere else in the same origin, should you continue those preloads and then transfer them to the document |
| 08:41 | <Noam Rosenthal> | ... I think it's useful to allow that, but not sure if that has further implications |
| 08:43 | <annevk> | Nothing comes to mind, but you might get rather involved sequences... 103 302 103 401 103 200 |
| 08:44 | <annevk> | Speaking of which, I guess those questions apply to 401/407 too. |
| 08:45 | <Noam Rosenthal> | yea I believe we'll have only the first or last (most likely first) 103 in the sequence apply |
| 08:45 | <annevk> | Ooh, I didn't realize that. Is that also according to the 103 spec? |
| 08:47 | <Noam Rosenthal> | annevk: no, we're discussing that as part of the HTML implementation of what 103 means |
| 08:48 | <Noam Rosenthal> | the RFC also allows sending 103 103 103 200 |
| 08:49 | <annevk> | Right, I kinda expected that would work, but I don't have a strong opinion I think. I asked someone in our networking team. |
| 08:49 | <Noam Rosenthal> | (working on a draft spec which would make this a bit easier to discuss) |
| 15:30 | <annevk> | TabAtkins: known that api.csswg.org is down? |
| 16:54 | <annevk> | Domenic: so I think that in Gecko in the "global view" for browsing contexts that have a parent it's not represented by a browsing context, but rather a specific Window (called a WindowContext), which unless we make that explicit would be more clearly represented by an element than what we have today I think |
| 16:55 | <Domenic> | Hmm. |
| 16:55 | <Domenic> | I think Chromium is the opposite, but I really need to think about the specific bad case you mention... |
| 17:00 | <Domenic> | annevk: if you could double-check my reasoning on https://github.com/whatwg/html/issues/7676 that would be very appreciated; if it's correct I'll need to do some serious surgery as part of the session history rewrite, but if I'm missing something and the current setup is OK that'd be great to know first. |
| 17:30 | <annevk> | Domenic: so having read that I'm not entirely sure where we end up with the wrong state |
| 17:30 | <annevk> | Domenic: the initial about:blank is wasteful, but it doesn't seem wrong |
| 17:30 | <Domenic> | annevk: the creation URL being about:blank seems most worrying to me |
| 17:30 | <Domenic> | Or is that a normal thing that happens? |
| 17:30 | <Domenic> | But I thought we did like security stuff with creation URL |
| 17:31 | <annevk> | Domenic: we do, but only when we're loading something, and by that time the creation URL will be the "final" URL, no? |
| 17:31 | <Domenic> | Does something later update creation URL? |
| 17:32 | <Domenic> | I thought it would only be set on creation... maybe that is the wrong assumption. |
| 17:32 | <annevk> | Domenic: as in, we'll create a new environment for the URL this new BC is being navigated to (that will replace the initial about:blank environment) |
| 17:32 | <Domenic> | I don't think we will? Because we have an initial about:blank so we'll just reuse the newly-created Window/ESO |
| 17:33 | <annevk> | No, it'll be cross-origin so it cannot reuse. |
| 17:33 | <Domenic> | Ohhh |
| 17:33 | <annevk> | (Creation URL is used by Fetch and is indeed security sensitive, but it's only used when the thing would fetch stuff, which this won't.) |
| 17:34 | <Domenic> | OK then yes I think everything is great. So glad I asked. |
| 17:34 | <Domenic> | I will add a big <p class="note">! |
| 17:34 | <annevk> | Fair 🙂 |
| 17:35 | <Domenic> | bakkot: logs are not working anymore? :( https://matrixlogs.bakkot.com/WHATWG/ |
| 17:56 | <bakkot> | Domenic: hmmm |
| 17:56 | <bakkot> | odd, they're working for other channels |
| 17:56 | <bakkot> | I'll take a look later today, thanks for the ping |
| 18:09 | <bakkot> | Domenic: should be fixed, sorry about that |
| 18:29 | <Domenic> | Awesome, thanks! |
| 18:55 | <TabAtkins> | annevk: Looks like the server went down real hard; I can't access the control panels either. Pinged plinss about it. |
| 23:18 | <shu> | Domenic: who should i be tagging for webidl PR reviews? |
| 23:18 | <shu> | (or anyone) |
| 23:19 | <Domenic> | shu: timothygu is good but busy, same as me; this week I'm in particular not doing reviews |