Re: [css3-exclusions] Issue 15183

On 8/1/12 1:27 AM, "Florian Rivoal" <florianr@opera.com> wrote:

>On Tue, 31 Jul 2012 16:11:02 +0200, Alan Stearns <stearns@adobe.com>
>wrote:
>
>> We are introducing an exclusion system that is orthogonal to positioning
>> schemes. Collisions are actually something the draft takes into account.
>
>Sorry for the poor phrasing. I did not mean that the exclusion spec
>was orthogonal to dealing with collisions, I meant it was orthogonal to
>avoiding collisions.
>
>> Please provide an example - of a layout that works fine in multiple
>> environments without exclusions, but with a simple exclusion causes the
>> layout to work in some environments and break in others.
>
>I don't have such an example, but you're missing my point.

I do get your point, I'm just continuing to argue for my version of your
option (1) below. Please understand that I do appreciate your responses -
continuing this conversation is very useful. I'm not trying to stop the
conversation by asking for examples. I think that examples of what we're
concerned about will help us decide on the way forward.

>
>I am not saying that exclusions make abspos any more broken than
>it already is. I am saying that exclusions, because they are
>useful and work fine on abspos, will result in a higher
>usage of abspos than we currently see.
>
>Given that Adobe in particular is pushing for this proposal, I am rather
>convinced that we will before long see GUI editors for web pages
>supporting exclusions (don't get me wrong, GUI editors are a good thing).

Here's what I think is a relevant parallel argument without referencing
exclusions. A GUI editor for web pages can use @media rules to output N+X
abspos layouts for N screens (X being how many screens can be set
landscape or portrait). Given this fact, you could argue that @media rules
result in a higher usage of abspos. What I conclude from this is that
those tools are bad, not that @media rules should deal with the problems
of absolute positioning.

>
>If you have a GUI editor for web pages that supports absolute positioning
>and exclusions, the temptation to create a box, position it somewhere
>manually, and tick the "flow around me" check box is great. So great that
>people might even use it in places where a float would have worked just
>fine.

And if you have a single abspos element with an exclusion (which I think
will be the vast majority of the cases, tool-based or not) then I do not
see a problem. I also do not see a problem with a small number of abspos
exclusions in continuous media, but there may be an example that shows me
I'm wrong. That's why I am asking for examples - to see the specific
problems we should consider relevant as we decide the way forward.

>...
>I see only 3 possible ways forward:
>1) accept a higher rate of abspos induced breakage
>2) reject any exclusion scheme that works with abspos and would tempt
>      authors into using it
>3) improve abspos first

If it would be acceptable to update the issue text for 15183 with these
choices, I would be very happy. As I mentioned in Hamburg, we have people
reading the spec then coming to me and saying, "I thought I understood
exclusions, but here it says they're based on absolute positioning, and
now it does not make sense." I would welcome a change in the issue text
that clarified the technical issues, and offered alternatives to discuss.

I would phrase option (1) as "accept that exclusions could encourage
abspos usage, and therefore make abspos problems more apparent" but would
argue that the benefits gained from the feature outweigh that downside (as
I believe is the case with @media rules).

>
>Many in this group do not want 1. I am sure that 2 does not appeal to you
>in the least, and it doesn't to me either. So I am trying to propose 3.
>
>I want collision:avoid on by default on abspos and fixpos which have
>exclusions turned on, and that means introducing the collision property
>before we introduce exclusions.
>
>Ideally, I'd want collision:avoid on on abspos and fixpos even without
>exclusions, but we can't do that without breaking compatibility.
>
>Introducing the collision property separately from exclusions is
>definitely possible, but then we couldn't have collision:auto compute
>to collision:avoid on abspos which have exclusions, as it would break
>compatibility. It would still be useful, as authors could opt into
>sensible behavior, but they would not have it by default, and I
>think that would be a loss.
>
>    - Florian
>

Received on Wednesday, 1 August 2012 16:01:34 UTC