[css-fonts-3] Default feature list should not require a list of features to turn on

[litherum]

https://drafts.csswg.org/css-fonts-3/#default-features

The set of default features should be able to be different on different platforms. Following platforms' text conventions is valuable.

[litherum]

cc @svgeesus

[svgeesus]

The Opentype feature registry uses should, not must. See for example
https://www.microsoft.com/typography/otspec/features_ko.htm#locl
"UI suggestion: This feature should be active by default."

So maybe we want to be more prescriptive for CSS than is required by Unicode, but the WG should discuss.

[svgeesus]

This is on the agenda for CSS WG telcon 13 Dec 2017. If we keep the must, I volunteer to write the tests for that section. But we would also want to put it at-risk because I don't think we have interop there. If we drop down to may, it becomes untestable (except for the 'switching off' clause). I would like to get this section updated either way, for an early 2018 updated CR,

[dbaron]

Is the proposal to add the allowance that additional features may be enabled by default (which seems reasonable to me), or are you proposing that some or all of the features currently required by that section not be required (I'd be more skeptical of this change, although maybe there are some specific ones that could change)?

[svgeesus]

I committed some tests for this section (each has several subtests). I believe these test everything that this section mandates. Here are the results:
https://test.csswg.org/harness/results/css-fonts-3_dev/grouped/section/7.1/

Firefox and Chrome pass all tests, so we do have interop there. Safari passes most subtests but not all, which does lead to failing each test it t's entirety. Edge does not currently implement this so naturally fails all the tests.

I agree with David that allowing implementations to exceed the requirements (can do better) is noble and also testable and workable. Watering down the spec to may is unfortunate for content developers, because they then have to set a bunch of stuff just to get interop between browsers.

[dbaron]

@litherum: #1267 (comment) is a question for you, really

[css-meeting-bot]

The Working Group just discussed css-fonts-3] Default feature list should not require a list of features to turn on.

The full IRC log of that discussion

<dael> Topic: css-fonts-3] Default feature list should not require a list of features to turn on
<dael> github: https://github.com//issues/1267
<dael> Chris: There's a set of features that should be turned on. A should not a must. dbaron commented saying he's fine with additional enabled by default but didn't want to relax. I made tests and sub tests for that section. FF and Chrome pass all. Safari passes most sub tests. Edge doesn't impl this at all.
<dael> Chris: My feeling here is if we water this down we're forcing content authors to stick a block of stuff to turn it back on if they care about doing what open type spec says you should.
<dael> dbaron: I'd like to know what myles wanted.
<dael> astearns: One of the bits in the comment is a claim things can be different for different platforms. Did you test that Chris ?
<dael> Chris: I did. I did mostly Win 10 and then Safari on Mac OS. I think it would be good to test on Andriod and IOS.
<dael> astearns: And FF and Chrome on Windows and MacOS
<dael> Chris: Right. I'm happy to test more.
<dael> dbaron: I'd still like to hear myles' answer for what he prop the spec change to do
<dael> Chris: I do, but I think astearns is saying I need more data too.
<dael> astearns: dbaron can you tag myles and ask that Q in GH?
<dael> dbaron: Yes.