[CSSWG] Minutes Telecon 2019-05-29 [css-sizing] [css-color-adjust-1]

=========================================
  These are the official CSSWG minutes.
  Unless you're correcting the minutes,
 Please respond by starting a new thread
   with an appropriate subject line.
=========================================


CSS Sizing
----------

  - RESOLVED: Rename Box Intrinsic Sizing to Box Sizing with no change
              in shortname (Issue #3962: Rename spec to CSS Box Sizing)

CSS Color Adjust
----------------

  - Allowing pages to say that they work for any color scheme was
      proposed for future-proofing, but the concern is that authors
      will use it assuming the only options are light and dark and
      when/if a third color scheme is introduced it will cause
      breakage.
  - RESOLVED: Reject this issue (Issue #3859: Let pages opt in to any
              scheme)
  - Adding sepia as a third color scheme (Issue #3853: Possible future
      <custom-ident> value 'sepia') received initial push back on if
      the use case is strong enough. Through discussing it was decided
      that allowing the top level color schemes to have sub-types
      accessible via the media query could be a way forward to add a
      users preference for sepia web pages but still using the light
      form controls. This idea will be explored further in the F2F.
  - RESOLVED: Keep this language (to allow a page to express a theme
              preference) in the spec (Issue #3850: Allow a page to
              express a theme pref when the user doesn't)
  - RESOLVED: Add color-scheme: normal (Issue #3849: Add
              'color-scheme: auto'?)

===== FULL MINUTES BELOW ======

Agenda: https://lists.w3.org/Archives/Public/www-style/2019May/0022.html

Present:
  Rossen Atanassov
  Tab Atkins
  David Baron
  Amelia Bellamy-Royds
  Oriol Brufau
  Tantek Çelik
  Dave Cramer
  Benjamin De Cock
  Elika Etemad
  Simon Fraser
  Dael Jackson
  Brian Kardell
  Rune Lillesveen
  Melanie Richards
  Florian Rivoal
  Jen Simmons

Regrets:
  Rachel Andrew
  Christian Biesinger
  Manuel Rego Casasnovas
  Alan Stearns

Scribe: dael

  Rossen: We can get going
  Rossen: Good day where ever you are
  Rossen: I wanted to call for any other agenda items or topics before
          we get going

CSS Sizing
==========

Rename spec to CSS Box Sizing
-----------------------------
  github: https://github.com/w3c/csswg-drafts/issues/3962

  fantasai: When we first drafted this spec it was limited but now it
            has all box sizing so should rename
  <florian> +1
  Rossen: Reasonable. Other opinions?
  AmeliaBR: No change to shortname?
  fantasai: Still css-sizing

  dbaron: A little worried about confusion with box sizing property.
  fantasai: Which is in this spec.
  dbaron: But it's a lot more.
  fantasai: Happy for other suggestions, but need to remove intrinsic
            and extrinsic from title
  AmeliaBR: CSS sizing too general?
  Rossen: I think so.
  fantasai: Box Alignment and Box Sizing go together
  Rossen: I think box sizing is great. Point about confusion is well
          taken, but property is in the spec. I don't think it will
          cause too much confusion
  Rossen: Unless there's better names lets resolve

  Rossen: Objections to rename Box Intrinsic Sizing to Box Sizing with
          no change in shortname?

  RESOLVED: Rename Box Intrinsic Sizing to Box Sizing with no change
            in shortname

How should inline-axis intrinsic sizing work for 'fit-content' and
    'fit-content()'?
------------------------------------------------------------------
  github: https://github.com/w3c/csswg-drafts/issues/3731

  fantasai: I haven't drawn up proposed edits. That needs to be done.
  fantasai: Unless someone has comments on general direction of edits
            we can come back once there's spec text
  Rossen: Remove from agenda for now?
  fantasai: That's what I'm thinking unless there's something to add.
  Rossen: It had been on the agenda for a few weeks. We can put it
          back to github only
  fantasai: Seems like in last discussion we were going toward oriol
            proposal
  fantasai: What I would do is edit that in. If anyone disagrees with
            that direction we can talk about that.
  fantasai: Otherwise I'll try to edit that in and see where we get.
  Rossen: Reasonable

CSS Color Adjust
================

Let pages opt in to any scheme
------------------------------
  github: https://github.com/w3c/csswg-drafts/issues/3859

  TabAtkins: Request is that right now a page has to explicitly say
             which schemes it supports. Only 2 but if we add more and
             a page is written for just light and dark and we added
             sepia you wouldn't get sepia on the page. Request is to
             say I support any color scheme.
  TabAtkins: Will work if you're careful to only use system colors.
             Won't work if you're not careful and you design for just
             light and dark.
  TabAtkins: Somewhat disinclined to add a new value that is opt me
             into future stuff. It's possible, but hard and rare.
             People don't use system colors usually because generally
             makes pages look bad. Most people using this won't have
             pages that look good. It'll be misused for just light and
             dark and sepia will break.
  <tantek> As someone who has recently added a few different color
           themes (dark etc.) to their site, I can agree with
           TabAtkins that this is nontrivial to plan for especially in
           the abstract!
  AmeliaBR: My argument to the contrary- Just because authors can be
            careless doesn't mean we should prevent authors from being
            helpful and user friendly. color-scheme: any is about
            being able to say as an author I'm lightweight with
            styling. I think that's a good thing that should be
            encouraged.
  AmeliaBR: Spec allows people to say which color schemes they have
            tested against. Even that we don't narrowly define what
            dark and light mean. A future browser could create a
            variation on dark and light that a careless website has
            problem.
  TabAtkins: Reason I'm still on my position is we want to allow
             authors to do the right thing, but also not design a
             feature easy to use wrong. Only way to design to take any
             color scheme is only system colors and almost no pages do
             that today. Perhaps we'll see that more in the future and
             then I'd reevaluate. The good practice to make this work
             isn't used right now so I'm not happy to add.
  dbaron: One slightly more common practice is don't use colors at all
  AmeliaBR: Simplest case is a lightweight website, a bit of spacing
            and layout but mostly trying to present text. In that case
            you want to say draw in colors user wants

  Rossen: How is this difference than not having at all?
  TabAtkins: Not having is white background black text. We don't have
             a way of saying this is unstyled
  Rossen: How is color-scheme:any different then not having MQ at all?
  AmeliaBR: Not a MQ. Property that tells browser which UA stylesheet
            to use. This is about if you use dark form elements or
            light form elements. Not a MQ.
  Rossen: Okay

  TabAtkins: Going for unstyled not using colors, it happens more than
             using system colors, but not a lot overall. I'd rather
             that's something that opts you in. If this was paired
             with a restriction that makes it so you can only use safe
             colors or no colors I'm fine. An uncolored page is rare.
             Without something that forces you into acting good I
             don't like something that will likely break in the future
             even if it works well today.
  <florian> I'm with Tab on this one
  <fantasai> I'm also with Tab on this one
  TabAtkins: Today it's a shorter way to write what they want but it
             breaks in the future when we introduce something new
  AmeliaBR: Seems backwards. Future focus is if I write a good coded
            website today it should work with future features.
  AmeliaBR: I would support adding an example or warning in the spec
            with a yellow on blue color scheme as a future example to
            make it clear that things might be different than what you
            expect
  <tantek> Worse than a footgun, this is timebomb
  florian: In general we should not assume people are incompetent.
           Lazy is reasonable. If they have something that appears to
           do the right thing today and when it's extended it does the
           wrong thing you could know that from the spec, but you
           didn't read the spec and it does what you want. I'm with
           TabAtkins on that
  TabAtkins: If there's a control that forces the good mode then I'm
             totally happy and we should have that option come in.
  AmeliaBR: I don't agree that's a necessary condition. Necessary
            condition is when you set font you also set background.
  <tantek> I'm particularly swayed by TabAtkins point that no one is (
           we have no examples of) using CSS that could take advantage
           of this.
  TabAtkins: If you have one element white bg and dark text with a
             dark color scheme you're not honoring intent of color
             scheme.
  TabAtkins: This is also generalizing before having evidence we're
             going to generalize. We right now don't have indication
             we'll introduce a 3rd color scheme. When we do I'd be
             happy to work on an any value. Right now no guarantee
             we'll introduce a 3rd. Giving an any now that would make
             it hard to extend I'm not comfortable
  <tantek> 1. no evidence (of anyone using CSS designs that would work
           with this feature), 2. no evidence of need for this value,
           3. likely to be a footgun or timebomb (worse)
  <fantasai> +1 to tantek
  <florian> +1 to tantek too

  AmeliaBR: IRC most people agree with you. I'll accept this as an in
            future if it's necessary but maybe not now. We will
            discuss a 3rd soon with sepia
  Rossen: Proposal is to reject this issue
  Rossen: Objections?

  RESOLVED: Reject this issue

Possible future <custom-ident> value 'sepia'
--------------------------------------------
  github: https://github.com/w3c/csswg-drafts/issues/3853

  AmeliaBR: Brought up fairly early on
  AmeliaBR: If we're talking about other color schemes a likely
            example is sepia
  AmeliaBR: afaik every browser with a reading mode has a sepia color
            scheme with a creamy beige background and dark gray text
  AmeliaBR: Color scheme is popular based on readability studies
  AmeliaBR: Also used on some websites as their own style, like
            Financial Times
  AmeliaBR: One thing missing is we don't have any UA or OSs with
            sepia scheme form elements and widgets
  florian: Not sure, but I suspect these exist in ebook readers
  dauwhe: Really common in ebook, but I don't think many have form
          controls
  <dbaron> It's not clear to me why this should be distinguished from
           'light'.

  myles: Reader mode in safari is picky about styles it lets through
         to reader mode. Do others work the same or do they dump the
         styles into reader mode?
  AmeliaBR: I think same in that they override most styles to apply
            reader mode color scheme
  AmeliaBR: We're talking about creating an option so on the main page
            a website could opt into those defaults so if user prefers
            them it can be on top. Wouldn't change which styles get
            into reader mode
  <dbaron> I don't know how Firefox reader mode works.
  myles: I'm trying to understand difference between a regular page
         like FT with a sepia scheme compared to reader mode where
         browser supplies most of the styles. Neither use case needs
         this

  <dbaron> This sounds like making up use cases.
  AmeliaBR: Use case is responding to user preference. Being able to
            have a named color scheme that you could ask prefers sepia
            from prefers light. They're different in terms of reader
            experiences.
  AmeliaBR: On your website using color scheme property being able to
            say these styles work with light or sepia, whichever user
            prefers
  TabAtkins: myles point is arbitrary webpages where users can want to
             view light or dark. Arbitrary pages is sepia seems a lot
             smaller. Reader mode replaces everything already. I
             haven't seen reader systems that do substantial CSS
             styling.
  TabAtkins: Those being specialized systems can do overrides on their
             own. Don't need arbitrary webpages to sepia
  florian: If sepia form controls are a thing FT might want to opt in
  <tantek> colorizing form controls is the easiest thing to customize
           about form controls

  florian: AmeliaBR you've been saying this controls form controls and
           default colors. Does it actually do that?
  TabAtkins: Supposed to
  AmeliaBR: florian's question, yes. If you set a color scheme on the
            root element then default background and text color should
            match that scheme.
  florian: Thanks

  dbaron: I still don't see why this is different from light. These
          schemes in reader mode are generally light colored. I don't
          see reason why not same as light.
  AmeliaBR: I brought up reader as an example where this scheme is
            popular. Readers and ebooks offer sepia distinct from
            light so it means readers find these things different.
  AmeliaBR: As a reader if I had a preference on a website to use
            sepia I would use that. Given it's popular in other
            reading environments I'm not the only one.
  AmeliaBR: Form control maybe light makes sense to use in both. But
            it's knowing difference between light and sepia
  dbaron: You could have a light with sepia-ish settings
  TabAtkins: Or a black on white and lighter and both trigger from
             white
  <tantek> one possible difference between "light reader" and "sepia
           reader" is treatment of color images - I'd expect the
           latter to sepia-ize images, but not the former
  <tantek> images, thus background images, decorative images etc.
  <AmeliaBR> tantek, I don't think that happens. This is about
             typography colors.
  AmeliaBR: How would an author respond?
  TabAtkins: MQ can offer finer grained. If you want to address sepia
             tones when can provide that so a page can opt into that.
  TabAtkins: Agree with dbaron now, it's a good point, need an
             argument why color scheme property needs more than those
             two. It's generally light or generally dark. Page doesn't
             know if it's high contrast light or sepia light.
  AmeliaBR: Good argument. Add color scheme in the media query but
            consider it light in color scheme?
  TabAtkins: Yes. Light and dark are super categories. As we ID useful
             others they're sub categories we add to the MQ only
  AmeliaBR: It's a practical solution that addresses use cases.
            Rethinking how property and MQ relate is required. Need to
            start thinking what the subsets mean for the MQ. What does
            it mean to prefer sepia but if it's not available light is
            better than dark. Ranking might come in
  AmeliaBR: Maybe try and find some time at F2F to think that it might
            look like
  AmeliaBR: Having prefers color scheme for sepia is the biggest
            advantage. Sepia form elements aren't a high priority for
            me.
  <dbaron> I'm still not convinced about the need to add this to the
           media query, at least as a keyword for the existing one.
           Maybe media queries against characteristics of system
           colors?

  dauwhe: This is a problem ebook has facer to balance user preference
          with content authors desire. epubs do a bunch of hacks. Even
          though these sound difficult I'm happy we're trying to talk
          about them.
  <dauwhe> https://github.com/readium/readium-css/blob/develop/css/src/modules/ReadiumCSS-sepia_mode.css
  jensimmons: Mostly have questions, but I can ask at F2F
  jensimmons: We'll figure out the right balance between not
              over-engineering and giving needed power. Open questions
              on how user of webpage spec what they want and will
              browsers provide those controls. Will designers design
              for many color systems? I'll ask these at F2F.

  Rossen: florian can we postpone to F2F?
  florian: I thought F2F was MQ. My question is when we set property
           on root and it opts you into the correct color. If you say
           dark you get dark but when that means different colors you
           often have different foreground and background. But then
           you have image with background it looks bad in sepia. If
           you say light you might not be fine with any light, you'd
           look bad with sepia.
  TabAtkins: I think that's exactly the deal we'll go with. Need to
             craft language. If you need white them spec white, if
             you're fine with anything light then use color scheme and
             you're good to good. Right now you have no way of knowing
             what type of light. It could be white, off
             white...problem exists today. You should set an explicit
             color if you need one
  <jensimmons> Web designers are not going to let go of control to the
               extent that ya'll are imaging they will.... things to
               debate about this over dinner & such next week.
  AmeliaBR: Getting into last issue, if people test in current
            browsers light mode has a white background and people
            aren't planning for future. Need to warn people what is
            and isn't guaranteed in color scheme
  <tantek> iOS has grayscale, red/green filter (Protanopia), green/red
           filter (Deuteranopia), blue/yellow filter (Tritanopia), and
           "color tint"
  <tantek> in the "Color Filters" setting inside "Display
           Accommodations" inside "Accessibility" inside "General" in
           "Settings"
  <dbaron> FWIW, browser default color backgrounds used to be gray
           rather than white, but basically all sites set a white
           background, so browsers changed...

  Rossen: Current issue as proposed I'm hearing pushback.
  TabAtkins: There were just more details. I'll write up in issue and
             we'll discuss at F2F
  Rossen: I'll change label to F2F agenda

Allow a page to express a theme pref when the user doesn't
----------------------------------------------------------
  github: https://github.com/w3c/csswg-drafts/issues/3850

  <jensimmons> There's a philosophical debate underneath all these
               details, about how much visual display could/should be
               under the control of the user vs browser vs author. 25
               years of evolution of the web has already answered a
               lot of this.... the web started expecting 100% browser
               / user. We ended up at 100% author. Light/mode will
               change this... but imho, only so far.

  TabAtkins: According to early proposal from rune color scheme can be
             used in 2 ways, what you're okay with the user selecting
             and say what you must be opted into. If you say only dark
             you get dark.
  TabAtkins: Middle ground is if they user doesn't have a preference
             please opt into this color scheme. This way pages that
             want to be dark can get dark without forcing the user
             into dark.
  TabAtkins: Proposal is we interpret color scheme without the only
             keyword we opt the page into the first in the list if
             they user doesn't have a setting.
  AmeliaBR: This is how it's defined in the first pass draft?
  dbaron: How does this distinguish between user wants it this way?
  fantasai: Current state of the web is user doesn't have a strong
            preference. Need to be careful not to mix preference for
            light with no preference
  florian: That's my problem with the property. OSX just has light and
           dark and so we end up with the same problem as any. People
           could order randomly and then when the OS gets a preference
           they get light first and they didn't know order matters
  TabAtkins: That's fine. If written to accept either user preference
             when this chooses your page will respond correctly. You
             might not have expected dark when you wrote dark light,
             but it works with dark
  florian: I guess

  tantek: I like this a lot. Lots of pages have theme to go dark or
          light. Triggered by the user, but lets us communicate to UA
          that this site is now choosing dark theme, do the dark theme
          other things.
  tantek: Because page has to iterate the list it takes explicit
          design. Less footgun
  florian: May have unintended but not breaking
  AmeliaBR: florian original point is a limitation of browsers that
            they don't have no preference. That's a good point. We
            have spec text encouraging browsers to have no preference.
            I don't think there's a concern about breaking anything.
            By definition author has said can do dark or light
  dbaron: I still think no preference is a complicated UI no one will
          try and get users to express. You can flip between users
          that look like this or like that. I don't picture someone
          building a UI for that
  florian: I like reddit being dark on light. I don't want them to
           support. I like supporting site rendering.
  <florian> I like github being light, and arstechnica being dark.
            These are there branding, and I have no desired to
            override everything into being all light or all dark
  TabAtkins: If it turns out no preference you don't use that part of
             algorithm. Once you do support no preference you already
             said you're okay with any. This prevents a browser that
             wants to support no preference from forcing everything to
             light
  <jensimmons> +1 to what dbaron said.
  <jensimmons> We aren't going to have a complex set of web browser
               preferences to set a pref for each website, one by one.
  <AmeliaBR> jen: Why wouldn't we have preferences for each website?
             Most browsers remember my zoom level preference for each
             site & allow me to easily change it. Why not colors?

  fantasai: Current state of the web is no preference. If we want
            preferences to be an enhancement then we need to have a
            default setting that is no preference so there's no
            degradation. In the OS you can ask light or dark. You
            controls aren't your user preferences on the page. Adobe
            uses dark controls but drawn on light content.
  fantasai: Similar to rest of UI for an OS. You might want to have
            dark controls and a normal page. Current state, use case
            for dark controls is I'm in a dark room and want to switch
            everything. Light mode it's less likely you hate dark
            themes. Don't want to force websites to switch that off.
            As interpreted it's binary preferences it's not true we
            get only light or only dark for the user
  <dbaron> Isn't the current state of the web 'light' rather than no
           preference?
  <tantek> implicit 'light' is the same as no preference
  <tantek> so I agree more with fantasai that current default is no
           preference
  <fantasai> dbaron, in terms of form controls yes, but not in terms
             of media queries. and these two need to integrate
  <fantasai> dbaron, legacy pages won't specify 'color-scheme: light
             dark' so we will know that they need light form controls
             for compat
  tantek: Current UI there's either no preference or an explicit
          setting of dark. I don't think it's right to assume current
          setting is light mode.
  florian: Based on websites today agree. How UI work macOS has a
           button for light and for dark

  Rossen: Back to original issue- TabAtkins request has specific
          purpose. Not sure we're addressing that or getting close to
          accept/reject
  florian: I think we've established it's harmless and desirable.
           Haven't established it will work.
  AmeliaBR: About enabling UAs to have this option without conflicting
            spec
  myles: Which OSs have or may have no preference?
  AmeliaBR: Not about OS level. About a browser that can say let
            websites use whatever color scheme they want rather than
            use my OS color scheme.
  <jensimmons> Do we have any requests from browser to have more
               complex controls? (What AmeliaBR is saying right now.)
               Has any browser asked for such control yet?
  Rossen: True for form controls
  <fantasai> See also https://github.com/w3c/csswg-drafts/issues/3857
  tantek: And things like scrollbars. Right now if a web author wants
          a dark theme they have to go through every stylistic and
          carefully make it dark. Being able to say set a dark
          background and light text, hey browser can you do the rest
          for me is great.
  tantek: This is another way to let authors do the smart thing easier
          then do things awkwardly
  <jensimmons> What we really need to do is expose all the visual
               styling of form controls. Not just light/dark/foobar.
               So authors stop replacing form controls with JS + spans
               & divs.
  <fantasai> jensimmons, that's being worked on but it's a really hard
             problem

  TabAtkins: This is one line in the spec that says if user has no
             preference use the first author preference. If there is
             no option for no preference you skip that line. It's
             future compat with no appreciable increase in complexity
             to the spec. It's good to aim for in the future.
  <florian> I support keeping it in the spec
  <bkardell> seems good to keep as in the spec
  <dbaron> I think the spec text is ok, but I don't think it's going
           to do the things people imagine it will do.
  <AmeliaBR> +1
  TabAtkins: If people want to remove we can discuss at F2F. I'd like
             to resolve to keep as spec
  tantek: +1
  fantasai: +1
  <bkardell> +1
  <dauwhe> +1
  Rossen: Objections to keep this in the spec?

  RESOLVED: Keep this language in the spec

  <fantasai> https://github.com/w3c/csswg-drafts/issues/3857
  fantasai: light vs no preference has a separate issue. I flagged it
            as for the F2F

Add 'color-scheme: auto'?
-------------------------
  github: https://github.com/w3c/csswg-drafts/issues/3849

  fantasai: Original version had light as initial value. That's
            required for web compat. Some UAs don't render with light
            style and they don't need to be web compat. Proposal is
            use auto and say if you're a UA that needs web compat you
            need to use light. If not you can use dark.
  <florian> call it "default" rather than "auto", and then I'm good
            with it
  TabAtkins: This is distinct from any. This is they author saying I
             don't know what color schemes are and you get default
  AmeliaBR: Light schemes are allowed to be more diverse then default,
            especially if sepia is a light scheme. That's not web
            compat. Light default is assumed to be black on white.
            Worth distinguishing.
  AmeliaBR: Other thing is the keyword should make it clear this is
            default and not just auto in sense of do it smart
  TabAtkins: Perhaps none is right
  florian: I think default is good name
  TabAtkins: Shouldn't use default. That's a reserved global keyword
  florian: Okay
  fantasai: Normal is a possibility
  florian: Sure
  myles: Reason against legacy?
  fantasai: Not necessarily legacy
  florian: Sounds deprecated

  Rossen: Normal vs any. I hear sympathy to add this. Let's figure out
          name.
  <florian> normal
  <AmeliaBR> +1 to normal
  TabAtkins: No one is going any. Normal or auto
  * tantek could live with normal

  smfr: We don't want auto because WebKit thinks that means use what
        meta tag says
  [General confusion about WebKit property]

  Rossen: Sounds like it's 'normal'
  Rossen: We can try and resolve
  fantasai: Should we just add 'normal'?
  TabAtkins: Use 'normal' as the word, sure
  <futhark> +1 to 'normal'
  fantasai: We'll do 'normal' now and figure out weird WK stuff later

  Rossen: Proposal: Add color-scheme: normal
  Rossen: Objections?

  RESOLVED: Add color-scheme: normal

  Rossen: End of call. Next call is in a F2F format. Safe travels and
          I'll see you Tuesday!

Received on Wednesday, 29 May 2019 23:01:46 UTC