[css3-layout] comment on comments, and my perspective on requirements

I wanted to send a few comments on
http://www.w3.org/TR/2005/WD-css3-layout-20051215

First of all, I've been rather surprised at the number of detailed
comments on this draft recently.  The draft already says that:
  # The contents of this document are still highly experimental.
To expand on this a little:  there are basically two reasons a working
group might publish a working draft:
 (1) because the draft reflects group consensus, or
 (2) to put out some ideas that are not group consensus for discussion.
My memory of the group decision to publish the draft was that this was a
case of (2), and in hindsight I should have asked that that be
explicitly stated in the status section of the draft.

In particular, I was one of the group members who disagree with a number
of the fundamental ideas in the draft, and I'd like to explain why.

My main problem with this proposal is that it puts two different things
into a single feature:  a new box model, and a mechanism for reordering
content.  We've seen in the past that that is dangerous.  Authors often
use an inappropriate box model because of its ability to reorder
content:  many uses of floats and absolute positioning, where authors
chose to use those features so they could present their content in an
order different from the document order, respond very badly to changes
in viewport size or font size.  (I discussed this more in the paper I
presented at XTech [1].)  We shouldn't force authors to use a flexible
box model to make navigational content that is at the end of the
document appear at the top of the screen presentation of that document
if the presentation is otherwise best done with normal block and inline
layout.  I'd rather we developed two separate features:  a reordering
mechanism, and a flexible box model.  I think a good and easy-to-use
reordering mechanism would be useful with the existing CSS display
types.

I'm also concerned about a number of other things in the document.  The
inability to address the regions with selectors seems problematic.  I
think the need for some properties, such as 'overflow', is likely to be
quite common.  Some important features common in user interface layout
toolkits, such as flexibility starting from an intrinsic width (rather
than flexibility starting from 0 width) also seem to be missing and
quite desirable.

There's been a draft of the "XUL box model" floating around for a while,
partly implemented in both Mozilla and WebKit.  I prefer that as a more
fully-featured alternative for the flexible box model aspects of this
document.  Hopefully somebody will get a chance to finish it up sometime
not too far in the future (maybe even me, but don't count on it).

I also think that more consensus about the answers to the four questions
at the end of my paper [1] might help focus the discussion about
flexible layouts in CSS.

-David

[1] http://xtech06.usefulinc.com/schedule/paper/146

-- 
L. David Baron                                <URL: http://dbaron.org/ >
           Technical Lead, Layout & CSS, Mozilla Corporation