| 01:07 | <Domenic> | It's not? |
| 01:08 | <Domenic> | I must other specs a decent amount |
| 01:08 | <Domenic> | Web IDL does all over the place |
| 04:32 | <annevk> | Yeah, I think that’s fine |
| 05:45 | <musgravejw> | hi there, I may have missed a thread on this issue (if there was one), was this a feature that is being considered? https://github.com/whatwg/console/issues/110 |
| 06:34 | <annevk> | musgravejw: all discussion on that is linked from there I think |
| 06:35 | <annevk> | musgravejw: if you’re referring to a mailing list, we no longer really use those in the WHATWG |
| 16:54 | <jyasskin> | Domenic/annevk: Specifications say how software works, and MUST specifies a requirement on implementations, sometimes multiple classes of implementations. Another specification isn't an implementation; it's another set of constraints on implementations. That said, it's clear, so if MUST is how we do it, I'll use MUST. |
| 17:09 | <Domenic> | I don't think specifications only say how software works. Sometimes they say how other specifications work. |
| 17:09 | <Domenic> | Again, that's most of what Web IDL does. |
| 17:55 | <annevk> | jyasskin: a web developer is not an impl |
| 17:59 | <annevk> | I thought https://www.w3.org/TR/qaframe-spec/#what-conform would be a good reference here, but it doesn’t really cover this explicitly |
| 18:10 | <annevk> | Maybe we should put some rules in Infra |
| 18:12 | <jyasskin> | Domenic: WebIDL mostly says, ~"when another specification says X, the browser implementation must Y." But you're right that it uses 'must' a lot to constrain what the other specifications can say. |
| 18:13 | <jyasskin> | annevk: We generally 'must' the server software if we're going to constrain the web developer. And, of course, 'must'ing HTML content was a disaster. |
| 18:13 | <jyasskin> | +1 for describing this in infra or somewhere. |
| 18:28 | <Domenic> | I've just never heard this rule that must can only apply to implementations... |
| 20:04 | <annevk> | Yeah, wonder about the source for that too |
| 20:04 | <annevk> | RFCs sometimes make rules for specifications |
| 20:07 | <annevk> | Mixed feelings on HTML myself; validator is generally considered useful |
| 20:07 | <annevk> | Is the thinking that a validator would emerge without a standard? Similar to eslint? |
| 21:49 | <TabAtkins> | jyasskin: Yeah, I use MUST to constrain things other than impls semi-regularly. Some in the group don't think it's reasonable to MUST an author, but it's like, you can't enforce it anyway, might as well be clear rather than conspicuously avoiding certain language. |