| 07:07 | <annevk> | We really need to fix that somehow |
| 07:07 | <annevk> | Maybe we should give up on HTTP fixing it |
| 07:07 | <annevk> | Guess I can reach out one last time |
| 09:33 | <annevk> | JakeA: do you have a pointer to the FetchController idea? Sounds interesting |
| 09:34 | <annevk> | JakeA: we do want to expose more things such as priority adjustments and H2 push |
| 09:34 | <annevk> | JakeA: sounds like that might offer it |
| 09:36 | <JakeA> | annevk: not sure we have anything beyond "a function that exposes an instance" |
| 09:37 | <JakeA> | I wonder if we can just subclass the promise and add this stuff on the returned value |
| 09:37 | <annevk> | JakeA: ah okay, so just a callback on RequestInit |
| 09:37 | <JakeA> | Yeah |
| 09:38 | <annevk> | JakeA: I like promise subclass personally, feels natural to me, but Domenic was not a fan |
| 09:38 | <annevk> | JakeA: callback seems good too |
| 09:41 | <JakeA> | Domenic: ^ is subclassing promises still a no-no? We're looking at adding methods onto the thing fetch() returns |
| 10:11 | <JakeA> | annevk: we could put observables on there for h2 push too. |
| 10:11 | <JakeA> | Although that gives us a TC39 dependency again |
| 11:22 | <annevk> | JakeA: yeah, but H2 push is in no hurry |
| 11:23 | <annevk> | JakeA: adjusting priority seems good |
| 11:23 | <annevk> | JakeA: as something sooner rather than later |
| 11:24 | <annevk> | JakeA: and termination of course |
| 11:24 | <JakeA> | annevk: the more I think about it the more I want a promise subclass again. Or proxy then/catch to an underlying promise. Subclass would be preferable |
| 11:24 | <annevk> | JakeA: and cheap progress reports although I'm a little unclear if we have sufficient use cases for that one |
| 11:29 | <JakeA> | annevk: upload progress certainly |
| 11:30 | <JakeA> | I think streams can be used for download progress efficiently, but probably needs things like BYOB reader |
| 11:30 | <JakeA> | To avoid GC |
| 12:17 | <annevk> | JakeA: someone else suggested we offer some kind of transform stream that exposes progress, but is otherwise transparent so the UA can optimize around that |
| 12:28 | <JakeA> | annevk: if it takes a stream it'll have to take a content-length too |
| 12:28 | <JakeA> | But that means you can create a generic progress transform |
| 12:29 | <annevk> | JakeA: content-length could be on top in theory, but yeah, I suppose |