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.