Re: [CSSWG] Minutes and Resolutions 2010-02-24


We have some comments on the CSS WG's evaluation of 'image-fit: auto':

fantasai <> skreiv Fri, 26 Feb 2010 23:43:23  
>    wrt A new 'auto' behavior
>    Bert: I do not like it; I think we can do without it
>    Peter: how do we get the default behaviors without it
>    fantasai: We can say that we assign the box and the content "filler"  
> does
>              what ever it thinks is right
>    fantasai: using the model above, the content filler is given the size  
> of
>              the area to fill and it makes the decision on how to fill it

This was previously suggested in the image-fit thread [1]. Erik Dahlström  
and I object to this as it limits the feature needlessly (as filling the  
content box will have an effect only for bitmaps). Furthermore, it is  
confusing that 'image-fit: fill' does not actually mean 'fill' from the  
user's perspective. (Erik will post on behalf of the SVG WG separately.)

>    Sylvain: Would 'auto' be the default behavior then?
>    Answer: yes
>    dbaron: Because "object" is so hard to implement, perhpas we should  
> not
>            force that on every other kind of element
>    Sylvain and dbaron: auto should not be the default just because it is  
> good
>                        for "object"
>    Daniel: do you agree that a new "auto" value is needed?
>    <dbaron> I think <object> behavior might be a bunch of quirks... and
>             object isn't used very much for any of this.

David Baron and Sylvain Galineau fear that accepting the 'auto' value  
would impose <object>'s complexity on all other types of elements.  
However, the 'auto' value isn't imposing anything new, our proposed  
specification is simply acknowledging existing, underspecified complexity.

>    <dbaron> I think the right behavior for <object> might be to switch
>             implementations to doing 'fill'.
>    Sylvain and dbaron: no, we do not agree there is a need
>    RESOLVED: the proposal for a new "auto" value is not accepted

If David Baron is using fantasai's definition (so that "fill" simply means  
that the viewport is the same size as the content box) this doesn't really  
change anything; if he is saying that all content in <object> should  
ignore aspect ratio and fill the content box (like bitmaps do), it breaks  
backwards compatibility. The latter is a no-go for us.

In the thread discussing this, fantasai said that format-specific details  
should be left out of CSS. While Anne van Kesteren and I agree that that  
would be the ideal, existing UA behavior is highly format-dependent. If we  
underspecify behavior here we increase risk of divergent rendering.

However, that is a possible compromise: We could give up specifying the  
rendering behavior of 'auto' and simply leave 'auto' to mean 'as usual'.  
That way we retain backwards-compatibility and all implemented features  
(by allowing 'image-fit: fill' to have an effect), and we would not need  
to include format specifics in CSS. This is in essence what fantasai  
proposed that 'image-fit: fill' mean, but moves fantasai's functionality  
to 'auto' (and leaves our implementation of 'fill' as it is).

This means giving up the goal of specifying rendering properly (in  
'image-fit: auto'), but I guess that is not any worse than the current  
situation. It would also mean giving up the idea of CSS not specifying  
anything beyond the size of the content box ('image-fit: fill', step 6 in  
[1]). I believe this combines the best of fantasai's and our proposals.


Best regards,

Leif Arne Storset
Core Technology Developer, Opera Software
Oslo, Norway

Received on Wednesday, 3 March 2010 10:51:49 UTC