00:29
<devsnek>
has there ever been any discussion about the reviver function and response.json()
00:30
<devsnek>
I'm guessing response.text() + JSON.parse is just good enough
00:55
<Domenic>
devsnek: https://github.com/whatwg/fetch/issues/104
01:37
<devsnek>
Domenic: thanks for the link
01:38
<devsnek>
it looks like it's assumed there that streaming json on main thread < buffer text and parse?
02:04
<Domenic>
I'm not sure how to parse that sentence
02:53
<devsnek>
Domenic: it seems like it was assumed in that thread that streaming parsing on the main thread isn't worth it compared to building up a string with .text() and calling parse
02:53
<devsnek>
which is surprising to me but of course I'm not an expert on performance
03:08
<MikeSmith>
reading https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/bKn2DLuRdeU
03:08
<MikeSmith>
about <iframe loading="lazy">
03:09
<MikeSmith>
> the <img> portions of the spec have landed, whereas we dropped a ball on the <iframe> variant landing in the HTML spec …
03:09
<MikeSmith>
> some have suggested
03:09
<MikeSmith>
un-shipping the feature.
03:09
<MikeSmith>
Why?
03:10
<MikeSmith>
Where? Who?
03:11
<MikeSmith>
I realize the message also says it’s not gonna be un-shipped, but I don’t understand why anybody would have suggested it to begin with
04:03
Domenic
MikeSmith: first message of https://groups.google.com/a/chromium.org/forum/#!topic/blink-api-owners-discuss/JjETiVoN6l8 has yoav suggesting removing the feature as one course of action
04:03
<MikeSmith>
aha
04:04
<Domenic>
(did not mean to /me that, I guess there's some IRCCloud keyboard shortcut that automatically inserts a /me prefix...)
04:04
<Domenic>
devsnek: "streaming parsing on the main thread" isn't well-defined.
04:05
<Domenic>
https://groups.google.com/a/chromium.org/d/msg/blink-api-owners-discuss/JjETiVoN6l8/3JeXh6rjAAAJ has Addy Osmani also suggesting moving it back behind a flag
04:06
<devsnek>
my thinking was that with the json source proposal it might be worth taking another look
04:06
<devsnek>
I imagine parsing out bigints will be somewhat common
04:07
<MikeSmith>
Domenic: thanks!
04:07
<Domenic>
devsnek: I think step 1 would be someone proposing an actual processing model for streaming JSON that has well-defined semantics. So far nobody has. The closest is NLDJSON.
04:08
<Domenic>
s/NLDJSON/NDJSON
04:09
<devsnek>
I was thinking step 0 would be "does anyone actually want this"
04:09
<Domenic>
Right, well, it's not clear what "this" is
04:10
<devsnek>
well parsing a bunch of json that might be broken into pieces is not difficult
04:10
<Domenic>
In a way different than just buffering into text and parsing all at once?
04:10
<devsnek>
well since all the modern json parsers in engines are loop based instead of recursion based
04:10
<MikeSmith>
Domenic: I had actually just mis-took the message I cited as being about un-shipping img[loading]..
04:11
<devsnek>
it should be as simple as just calling into a function every time you get more data
04:11
<Domenic>
devsnek: you can't produce any incremental results because if you have `{ ... 500 KiB of valid object literal JSON ... ]` then the whole thing must be a parse error.
04:11
<devsnek>
ah I see what you mean
04:12
<devsnek>
yeah that is unfortunate I guess
04:12
<devsnek>
scrap everything I said
04:12
<Domenic>
You need a new format that can have "checkpoints" for validity. NDJSON is one stab at that, but I'm unsure on the adoption.
04:14
<devsnek>
ndjson would serve the use cases I know of but I wouldn't suggest it being part of browsers
04:41
<annevk>
https://siguza.github.io/psychicpaper/ (test your parsers or ideally just have one of them)
04:44
<annevk>
Oh lol, they mitigated with more parsers
04:47
<annevk>
But also, what I thought was usage of XML is in fact several terrible dialects of XML
05:22
<MikeSmith>
annevk: wow
05:22
<MikeSmith>
> We can also ask ourselves how a bug like that could ever exist. Why the hell there are 4 different plist parsers on iOS. Why we are still using XML even. But I figure those are more philosophical than technical in nature.
05:22
<MikeSmith>
haha
05:31
<annevk>
Another instance of technical debt not being prioritized
08:00
<nixi>
hi
08:01
<nixi>
i have been looking for a formal specifications on which codecs to use in html video and i am not sure there is any
08:02
<nixi>
i have ben searching here mainly: https://html.spec.whatwg.org/multipage/media.html#the-video-element
10:51
<annevk>
nixi: there are no mandated formats atm
10:55
<nixi>
annevk: ah ok, so it is a free-for-all in terms of browser implementation of formats
10:55
<annevk>
unfortunately, yes, though I think through WebRTC there might be some mandated formats, but those might be scoped to WebRTC specifically, not sure
10:56
<annevk>
Hopefully AV1 is the way out of this mess
11:07
<nixi>
no daala?
13:15
<annevk>
If anyone is interested in somewhat more formalized storage architecture: https://github.com/whatwg/storage/pull/86
14:15
<annevk>
Another thing, I noticed that our snapshots now use the snapshot logos and it's rather nice: https://whatpr.org/storage/86.html
15:12
<Domenic>
There's a bug with the warning that I can't repro locally... it should say "this is a PR preview" not "not ready for implementation". But it only shows up on whatpr.org, not when I run the same commands locally.
15:21
<annevk>
Domenic: I suspect the PR preview software might need changes; tobie added that warning before we had something iirc
15:21
<Domenic>
Hmm
15:23
<Domenic>
Ah, yeah... https://github.com/tobie/pr-preview/blob/a38138ef02f65d8829a2c01c5a7195fa7b4ea941/lib/models/config.js#L44-L47
15:23
<Domenic>
It overrides the default md-warning set by our boilerplate
15:23
<Domenic>
That's awkward
15:23
<Domenic>
Maybe we can hard-code a WHATWG exception
15:27
<Domenic>
Filed https://github.com/tobie/pr-preview/issues/55
15:29
<annevk>
We could branch on params["md-status"] I guess
15:30
<Domenic>
Yeah that's what I ended up suggesting
15:36
<annevk>
https://github.com/tobie/pr-preview/pull/56