04:43
<annevk>
timothygu: say you have schem://host.com and then set it to http, would you add a path? or you have blob:blah and set to http, what then? I forgot why the reverse is disallowed, might have made things easier for Location or perhaps just for consistency
04:46
<timothygu>
Is there anything bad about saying "yes" to adding a path? Though it's reasonable to require a host for the non-special URL (else it's a no-op)
04:53
<timothygu>
annevk: ^^
04:57
<annevk>
timothygu: note that the host would also have to be reparsed or some such, they're just not compatible
04:58
<timothygu>
annevk: oh right…
13:38
<hsivonen>
The HTML spec has explicit support for a character encoding menu, but only WebKit-based browsers still have it. I think we should just hand-wave it instead of working it into the algorithm. (If we work it into the algorigthm, I think we should write down the security characteristics of Gecko's implemenatation instead of writing down WebKit's approach.) Opinions?
13:40
<hsivonen>
In Gecko, as far as the algorithm is concerned, there is now a "force detector" flag.
16:27
<annevk>
smaug: Edgar Chen: timothygu: interesting IDL issue here: https://github.com/web-platform-tests/wpt/issues/31125. It seems that EventHandler should not accept non-callable objects (at least not without them getting wrapped), but in implementations it appears that it does. Am I missing something here? It's hard to believe this was overlooked for so long.
16:42
<Andreu Botella (he/they)>
I was just looking into that yesterday
16:55
<annevk>
Thanks. I'm still lost at how to read that IDL conversion thingie, but it does seem the other pieces add up at least
16:57
<Andreu Botella (he/they)>
The wording is tricky, but it seems to say that, in any case other than [LegacyTreatNonObjectAsNull], converting non-callable objects will throw
17:06
<annevk>
Yeah, it's coming back to me now. So nice that every five years to a decade you get to learn silly things anew.
17:23
<annevk>
Thanks a lot for the help!