- From: Tab Atkins Jr. <jackalmage@gmail.com>
- Date: Wed, 31 May 2023 11:50:20 -0700
- To: fantasai <fantasai.lists@inkedblade.net>
- Cc: "C. M. Sperberg-McQueen" <cmsmcq@blackmesatech.com>, spec-prod@w3.org
On Tue, May 30, 2023 at 6:57 PM fantasai <fantasai.lists@inkedblade.net> wrote: > On 5/30/23 20:03, C. M. Sperberg-McQueen wrote: > > The idea that all instances of the keywords should be read using RFC > > 2119 is an attractive one, but it does not often seem to work in > > practice. I have frequently heard WG members and editors claim that all > > occurrences of 'must' and 'may' are to read as related to conformance, > > but I do not remember ever seeing such claims confirmed by inspection: > > there was *always" some use of "must" which meant "it will in the nature > > of things always be the case that ...", or some use of "may" meaning "it > > could happen that ..." (When these are pointed out, those WG members > > who five minutes before were arguing that special formatting is not > > needed to identify the conformance-related usages of the word because > > ALL usages of the word are conformance related have now abandoned that > > claim and are suddenly arguing that special formatting is not needed > > because no one could possibly be confused by using the words in their > > ordinary-language sense as well as their conformance-related sense.) > > Afaict, the CSSWG, WHATWG, and at least some other groups consider such > instances to be editorial errors, and will make an effort to reword these > occurrences. Right, *all* uses of the 2119 words are meant to be interpreted as 2119 semantics in the CSSWG, and in anything using Infra. There are plenty of ways to reword English to avoid them when those semantics aren't desired; I feel it's *very* confusing if a "must" has semantic meaning in one place but not another (and it's especially bad if that is only expressed via visible styling). My plan for Bikeshed is to introduce a very minimal shorthand for explicitly stating that you mean the 2119 keyword, and then lint for any usages that aren't marked up as such, suggesting you either mark them up or reword. I just filed <https://github.com/speced/bikeshed/issues/2564> to track it, my plan is probably to make the text `2119may` expand into a marked-up "may", and similar for the others. ~TJ
Received on Wednesday, 31 May 2023 18:50:46 UTC