Re: Center DIV

> Simply that the margin:auto technique is no use unless the block to 
> which it is applied does not have the automatically computed width. At 

As far as I know, tables are "shrink wrapped" with margin: auto, although
only in the CSS2 errata.  If that is true, using display: table-cell
covers the situation.  However, Boris has pointed out that once you
are into this situation things are not properly specified, which probably
means it is very difficult to create a simple algorithm, which may be
one reason why such features are shied away from.

> This all comes back to my point about CSS being a useful language for 
> nailing things down by the pixel, but relatively tricky to use when all 

But nailing to pixels is a bad thing to do with it.

> you want to do is describe the broad relationships between objects and 
> have the browser do the work of putting them together in a way 
> meaningful to the particular client. Such an approach to design is very 

I believe that pre-HTML desk top publishing systems had this sort of
capability, but relied on the human user to provide input as to
what actually looked good.  What was then published was final form,
such as PDF.

My impression is that, even for the case where you can solve for the whole
document at once, creating fully automatic systems is still a research
topic, but HTML is designed to be incrementally rendered, throwing in the
added complication that any solution has to avoid diruptive re-rendering,
or deferred rendering (the latter is, even now, a problem created by
people who use table-layout: auto tables, the default, for layout).

I think it would be a good thing if such a language were possible, but
I'm not sure it should be seen as a version of CSS.  I'm afraid that most
authors would find the abstraction too difficult to handle.  Most people
don't use basic word processor formatting properly.

Received on Sunday, 6 July 2003 16:47:42 UTC