Remove font-variant @font-face descriptor from Fonts 4

[litherum]

WebKit is the only implementation, and our implementation has serious architectural-level issues. Given we are the only implementation, we should remove this descriptor from the spec (and remove the busted support from WebKit)

[faceless2]

Although our implementation isn't ready for public release yet, we do have full support for this, and we have recently added some test-cases to reflect this at web-platform-tests/wpt#9397, with accompanying discussion at #2265.

[css-meeting-bot]

The Working Group just discussed font-variant @font-face descriptor.

The full IRC log of that discussion

<dael> Topic: font-variant @font-face descriptor
<dael> github: https://github.com//issues/2531
<dael> myles: Is the commentor in the room?
<dael> ChrisL: It's Mike Gramford (sp?) and I don't know whose impl it refers to. They did contribute some tests so there's a test and that's where it was discovered the font-varient descriptor was under impl.
<dael> myles: It's mostly in the issue. Similar to the @font-face descriptors there's one that lets you turn on/off features. Nothing to do with font selection. I thought webkit was only impl, but there's some other behind a flag.
<ChrisL> test case for font-variant descriptor https://www.w3.org/People/chris/fwf/font-variant-descriptor-01.html
<dael> myles: Webkit is only shipping and our impl has archetectural problems and there's bifircation of text code paths and we make that before we inspect @font-face. we need to know if there are features to turn on before we decide. If you're on the wrong path it migh or might not work.
<dael> myles: I'd like to remove it.
<dael> astearns: Does anyone claim Mike?
<dael> TabAtkins: I think it's a pdf renderer
<dael> astearns: We could defer.
<dael> ChrisL: I posted this on www-style. This is one of the non-passing tests. Do we defer? Do we delete? It's holding us from rec
<dael> astearns: this ability seems useful.
<dael> ChrisL: I thought so too but myles talked about how it can conflict with cascade and if you put this in @font-face you have fixed positions
<dael> fantasai: property overrides descriptor so no cascade problem.
<dael> myles: physiologically features should be applied to everything in the cascade.
<dael> fantasai: There are fonts out there that are always small caps
<dael> myles: But turning it on on small case will be fine.
<ChrisL> s/physiologically/philosophically/
<dael> fantasai: But there are other features where if I have this font I want historical ligatures but not on this one because it's ugly. It's abigger issue for stylistic alternates.
<dael> astearns: We have a spec we want to push and a feature we don't ahve 2 impl I think we should defer and not remove until we talk.
<dael> astearns: Objections to defer font varient descriptor to L4?
<dael> resolved: defer font variant descriptor to L4

[astearns]

@faceless2 in the minutes above, “until we talk” means until we talk to you. We aren’t yet removing this, just postponing deciding whether it’s specified to the next module level.

[faceless2]

Thank you for your consideration! But no real need. All I wanted to add was that there was another implementation out there, as I'm aware that that can make a difference to the decision process. Yes, ours is (or will be) a PDF renderer, along the lines of Prince or PDF Reactor.

We do have a working implementation of this part of the spec as it stands, although (IMHO) the spec could be modified to better serve the purpose it was intended for - some of this was covered in the issue I referenced and I know everyone is aware of those.

To sum up: it doesn't bother us in the slightest deferring this until L4, please do.

[svgeesus]

I made an equivalence test for font-variant descriptor vs. property
https://www.w3.org/People/chris/fwf/font-variant-descriptor-01.html
https://www.w3.org/People/chris/fwf/font-variant-descriptor-01-ref.html

Now merged into WPT
web-platform-tests/wpt#10415
and in CSS WG test harness
https://test.csswg.org/harness/test/css-fonts-3_dev/single/font-variant-descriptor-01/

Chrome (Canary) Firefox (Nightly) fail as they don't implement the descriptor. Safari (Tech Preview) passes the descriptor subtests but fails one of the property subtests. Edge (17 preview) fails.

[svgeesus]

@litherum could you please review the trivial test changes which just re-assigns these two tests to Fonts 4. I checked that all the links correctly resolve in Fonts 4.

font-variant-05 is one of only two non-passing tests for Fonts 3 at this point.

[svgeesus]

This test got moved to the Fonts 4 testsuite. There are still zero passing implementations

[svgeesus]

@litherum besides the "remove from Fonts 3" component (already done) I also get the impression this issue is "should it be there at all" so I am leaving it open. Can you confirm?

[svgeesus]

Still no passing implementations for font-variant descriptor. Do we want to keep this? Is there implementer interest?

[drott]

At this point we don't implement other descriptors in @font-face, such as font-feature-settings and font-variation-settings - @litherum, do you suggest to keep those and only remove font-variant?

[litherum]

Let's remove them all, if possible!!! (For the same reason I listed in the top post here)

We may have compat problems, because we've been shipping the other ones for a while. I'm willing to try it, though! I have received bug reports about the font-feature-settings descriptor, so I know it's used at least somewhat.

[svgeesus]

@drott wrote

At this point we don't implement other descriptors in @font-face, such as font-feature-settings and font-variation-settings - @litherum, do you suggest to keep those and only remove font-variant?

I prefer to discuss those in a separate issue.

Edit: this issue

[css-meeting-bot]

The CSS Working Group just discussed Remove font-variant @font-face descriptor from Fonts 3, and agreed to the following:

• RESOLVED: Remove font-variant @font-face descriptor from Fonts 3

The full IRC log of that discussion

<mstange> Topic: Remove font-variant @font-face descriptor from Fonts 3
<mstange> github: https://github.com//issues/2531
<mstange> myles: We have two text paths. In WebKit. The distinction which path to go to has to happen before font fallback.
<mstange> ... Sometimes we detect something too late and can't go back and re-do everything, so we just do our best.
<mstange> ... If this is not necessary for the web platform, let's remove it.
<mstange> fantasai: It does look like there is an implementation (faceless2 says it in the issue)
<mstange> ... If the feature doesn't have a problem, we should leave it. We do not need implementations to move this spec at the moment.
<mstange> myles: When is the deadline? It's been years.
<mstange> Rossen_: This is a non-verifiable implementation. How do we evaluate it?
<Rossen_> q?
<mstange> drott: We would support removing it because the semantics are unclear in some cases.
<mstange> heycam: We would also be fine with removing it. We haven't heard much demand from authors.
<mstange> ... In principle it seems like a useful thing, but we would be fine with removing it.
<mstange> Rossen_: Any objections to removing it?
<mstange> Rossen_: No objections.
<mstange> RESOLVED: Remove font-variant @font-face descriptor from Fonts 3

[drott]

I prefer to discuss those in a separate issue.

Makes sense, let's track those separately.

[heycam]

This means we should also remove variant from the FontFace interface in the Font Loading API spec.

[litherum]

@heycam only edited the font loading spec, not the fonts spec.

[litherum]

Oh, I see, @svgeesus did it in d3db869 in fonts-4, and it was never in fonts-3 (meaning this issue is incorrectly titled).

[litherum]

Wow, it looks like I accidentally removed the relevant section from the "Feature and Variation Precedence" section in 0170cbc. Oops! Good thing we eventually deleted it anyway!