Re: 'border-image' confusion

On Jan 26, 2011, at 11:46 am, Brad Kemper wrote:

> On Jan 26, 2011, at 10:28 AM, "Tab Atkins Jr." <jackalmage@gmail.com> wrote:
> 
>> On Wed, Jan 26, 2011 at 10:19 AM, Eric A. Meyer <eric@meyerweb.com> wrote:
>>>  I suppose, but it seems like saying "100%" or "5px" once (and then being
>>> able to control how the image is repeated, with round/space/repeat) isn't
>>> all that more difficult than having a special keyword.
>>>  Is there some reason why the current behavior of 'border-image-slice' is
>>> desirable?  If there is, then yes, a new keyword would be needed to make the
>>> "use a single symbol all the way around" case happen.  But I don't see why
>>> slices overlapping should cause those slices to be forced to complete
>>> transparency in the first place, so I don't see why syntax changes are
>>> needed.
>> 
>> I don't think overlapping regions actually make sense.  You can define
>> an unambiguous handling of it, but it'll make about as much sense as
>> the location of the edges of an inline containing block. (In that
>> situation, you can have a "left" edge that's further right than the
>> "right" edge.)  This is why I think the current treatment of
>> overlapping regions is fine, and why I think it's clearer to simply
>> say directly that you want the whole thing, via a keyword rather than
>> lengths.
> 
> Yeah. I suggest we look into the "one tile and only one tile" border image idea for CSS4 borders. For that, you don't need any of the slice syntax, and can probably just stick the keyword in it's place (off the top of my head, I'm not sure how to handle the corners if the border widths are different). It was never a problem that CSS3 border-image was trying to solve. 
> 
> I'm also interested in adding 5x5 grids for border-image (in CSS4, so as to now slow down the CSS3 version), so that you could have pieces that are centered on each side, and tiles with images that pointed to the nearest corner. There are a lot of Art Nouveau border designs like that. 

I think there's also a valid use case for a single border-image that gets stretched, un-tiled, to the border box size, with the center clipped out. This would give us the ability to use a gradient image here, thus fulfilling the gradient border request.

Simon

Received on Sunday, 6 February 2011 18:12:59 UTC