[CSSWG] Minutes and Resolutions 2009-09-02

Summary:

   - Discussion of box-shadow and border-image.
     Brad actioned to come up with a rough proposal for all desired shadow
       functionality for comparison
     fantasai actioned to add more control to existing box-shadow through
       honoring background-clip and border-image-outset

   - RESOLVED: Republish Media Queries and CSS2.1 after Bert adds media_list
               grammar rule

   - CSS2.1 Issue 128 'display: run-in' clarifications bounced back to mailing
     list for discussion

   - fantasai to respond to UPnP forum message wrt status of CSS2.1

====== Full minutes below ======

Present:
   César Acebal
   Tab Atkins
   Bert Bos
   Elika Etemad
   Sylvain Galineau
   Daniel Glazman
   Brad Kemper
   Chris Lilley
   Hĺkon Wium Lie

<RRSAgent> logging to http://www.w3.org/2009/09/02-CSS-irc
Scribe: fantasai

* sylvaing speaking of boxes and diagrams, second Gerard Talbot's comment
            that the inline formatting model could use visual back-up
Some discussion of box-shadow
<fantasai> who's minuting?
<glazou> fantasai: that's outside of the call
Bert: I don't want to have a non-continuous border change the shape of the
       box. I still want the box to be rectangular.
<glazou> we did not start yet

Agenda
------

   Daniel: I'm back from summer break, not quite caught up with everything
   Daniel: Extra agenda item from Chris
   ChrisL: It's already in the Agenda

box-shadow and border-image
---------------------------

   ChrisL: ...
   ChrisL: You know the diagram with the border and padding edge etc?
   ChrisL: Imagine a similar diagram with the corners rounded
   ChrisL: I asked which diagram we use (?)
   <ChrisL> http://www.w3.org/Style/Group/css2-src/images/boxdim.png
   ChrisL: Bert said the boxes should be rectangular
   ChrisL: but you have a different clipping effect
   ChrisL: I also asked if the border-image affects the geometric layout
           of these boxes
   ChrisL: And I was told no, it doesn't
   ...
   Brad: So the question is, when is the border-radius applied and when
         is it not
   fantasai: The border-radius is applied to the definition of the boxes
             for rendering effects, but not for layout
   fantasai: for layout you only care about the edges, not the shape of
             the corners
   <TabAtkins> http://www.bradclicks.com/cssplay/curved-corner-image.html
   ChrisL: So there's no way to prevent text from leaking other than
           providing sufficient padding
   fantasai: right
   <sylvaing> CSS2 borders clip at the inner border edge; Brad's example
              clipped at the outer border edge
   ChrisL: was the rendering that Brad showed per spec or an implementation bug?
   fantasai: Implementation bug. The spec says to follow the curve for
             clipping effects, jumping from one curve to the other isn't
             a reasonable interpretation

   ChrisL: fantasai's suggesting this particular unioning algorithm
   ChrisL: I think it'll look a little odd. I can produce images that look
           good, and some that look bad with it
   ChrisL: You don't need to do edge detection
   ChrisL: As you rightly pointed out, spread is a complicated operation
           and if you have a raster image it's out of the question
   ChrisL: blur radius is fine, but spread basically moves the geometry
           out and that's difficult to do without actual geometry
   Brad: I disagree with the notion of doing alpha channel for box channel
          without taking into account border styles or backgrounds
   Brad: I don't see why border-image should have a special alpha-based
         channel behavior
   Brad: I'd rather see a proper drop-shadow effect rather than this
         half-effort for border-image
   ChrisL: It's hard to address pieces of the border image otherwise
   Brad: If we had a drop-shadow property, it could have a switch on it
         that says what it affects: borders, border-images, everything, etc.
   Brad: So far box-shadow only takes into account the border box
   ChrisL: it does take into account border-radius
   Brad: Still limited to flat edges and possibly curved corners
   Brad: If you have a dashed border going around the curve
   Brad: It's going to be a flat border until you put a border image
   * TabAtkins or hell, just an identical border-image of a dashed border.
   Brad: You'd get a different effect if you had a border-image shashed
         border vs a UA-defined dashed border
   <ChrisL> http://www.bradclicks.com/cssplay/border-image/Alladins_Lamp.png
   fantasai: Two things come to mind from this discussion
   fantasai: One is, we could use the background-clip on the bottommost
             background layer to determine whether the dashes fall inside
             the box-shadow's auto-opaque area or outside it
   fantasai: box-shadow is supposed to represent clipping the box out and
             pulling it forward of the canvas
   fantasai: if you filled it with background paint, then it's obvious
             that in the default case (background-clip: border-box) your
             box would be rectangular even in the case of dashed borders
   fantasai: but if you set background-clip: padding-box, the dashes are
             outside the background paint region
   fantasai: and if you were clipping out the box, you'd clip around the
             dashes, leaving gaps in between
   fantasai: the other thought was, Brad mentioned perhaps never
             alpha-channel-masking the border area
   fantasai: and if the author needed alpha-channel masking, it would only
             apply to regions of the border-image outset beyond the border-box
   Tab: If we're trying to be intelligent about alpha, we can't just pay
        attention to the bottommost background unless we say background is
        automatically opaque
   fantasai: we're saying that the background is automatically opaque
   Brad: The box shape that you're shadowing, it's not really paying
         attention to the alpha of the box it's just pretending it's opaque
   ...
   Brad thinks authors will want more controls over what is drop-shadowed
   fantasai: we can always add more controls later, and this would just be
             the default behavior
   Tab: Maye be don't need to make this behavior that intelligent then
   ChrisL: I was agreeing with fantasai, but now I see both designers not happy
   ChrisL: If we make it less intelligent, then you always get a rectangle
   Brad: Initially my feeling was to suppress the shadow, because of your
         argument that we're not changing the box shape we're just decorating it
   <Bert> (Maybe we can define shaped boxes some day, which shape the content
           box and hence the text, rather than just the border...)
   fantasai: How about you take an action item to draw up a proposal for what
             kind of controls we want in the future
   fantasai: so that we know whether box-shadow conflicts with that or adds to it
   SteveZ: Two comments. First, this is hard to follow without pictures
   <ChrisL> I'm happy to make diagrams, once we decide what to make a diagram *of*
   * TabAtkins hack possibility: <div style=width:0 margin:20px border-image:foo>
                                 <div style=margin:-20px /></div>
   <glazou> did I hear howcome on the call ?
   <glazou> who said "'I support Bert" ?
   SteveZ: Second, what I'm hearing is that Tab and Brad want to shine a light
           through the box, and take the shadow from that, and having part of
           the border-image take part in that and part of it not
   Brad: I like the idea of using backgrond-clip to determine which box gets
         shadowed
   Bert, Hakon: I think it makes it all too complicated
   ACTION: Brad to come up with full shadow proposal (not finalized, but to
           see where we're going)
   ACTION: fantasai draw up proposal for box-shadow modified by ideas posted here

Media Queries
-------------

   RESOLVED: Republish Media Queries and CSS 2.1 after Bert adds media_list
             grammar rule
   discussion of process requirements for republishing CRs

CSS2.1 Issue 128
----------------

   <glazou> http://lists.w3.org/Archives/Public/www-style/2009Jul/0025.html
   display: run-in clarifications
   Bert: We'd made some progress on the mailing list after some very good
         questions from bz
   Bert: So we know ... and where floating children are displayed
   Bert: What we don't have yet is what happens with :first-line and :first-letter
   Tab: I support Sylvain that :first-line/:first-letter still cause problems.
   Tab: Don't know for the positioning ancesters of an abspos child of the
        run-in.
   Bert: My position is that it's the element hierarchy rather than the box
         hierarchy
   Bert: That's how I interpret the rules in 10.1
   Tab: You should get in on the thread; bz is expressing the opposite
   fantasai: Sounds like we need more mailing list discussion

Message from UPnP forum
-----------------------

   <glazou> http://lists.w3.org/Archives/Public/www-style/2009Aug/0605.html
   Daniel: I just noticed this email yesterday
   Daniel: Apparently Toby wants an answer before today
   ACTION: fantasai respond to Toby
   SteveZ: You can invite them to contribute test cases
   ChrisL: I think asking them to help review test cases would be better :)
   SteveZ: ok, /contribute/ to the test suite process
   Meeting closed

Received on Wednesday, 9 September 2009 19:57:48 UTC