Designers who think that CSS is in any way fine as it is are—and I say this as kindly as possible—simply so resigned to its limitations that they have unconsciously lowered their standards and expectations to meet what CSS actually makes possible. The range of selectors it provides is passable, but incomparably poorer than alternatives; the complete lack of variables and sophisticated constraints, to name a few. The ALM is a big step in fixing one of these inadequacies, but in my opinion it doesn’t go nearly far enough. I’m suggesting XSLT not as a difficult new technology for neophytes to master, but as just one part of a possible alternative.

to paraphrase myself from elsewhere, if people are still stressing about clearing floats in 2020, we really haven’t been working hard enough. My suggestion was thoroughly speculative; I’m wide open to better alternatives, just so long as the problems of today are actually fixed this decade instead of being swept under the rug.

display:table will never be a realistic option, content source ordering is impossible, and on that point, not that far from using actual tables.

floats will be here for a long long long time, why? because CSSP has been one huge clusterfuck right from the start and no one at the W3C is little more than a slumbering dinosaur that seems incapable of producing a spec in under decade…

Also, while I acknowledge the satisfaction that achieving a high level of proficiency in CSS confers, the fact that it requires such effort—to the extent that to my knowledge, there are no visual editors which can produce CSS layouts from scratch with the same robustness and flexibility as existed for old-school table markup—rather highlights its limitations.

If I may propose something rather radical: after several years of considering this matter, I actually think that CSS is quite far from being an ideal method of presenting a semantic HTML document. I have three reasons for this belief:

1. CSS, at its most fundamental level, does not actually separate content from presentation. Any presentational effect in CSS is basically done by somehow “hooking” some combination of styles and positioning onto an element of content—effectively, the structure and number of elements in your markup limits what you can do with CSS , which is the biggest reason that so many websites are replete with non-semantic “wrapper” or “container” elements.

2. In order to achieve the above, CSS must first select the correct elements to style, in which respect it is also far from ideal—as anyone who has ever hoped for a parent or next-sibling selector will know.

3. Finally, as I hope any graphical designer will appreciate, CSS in and of itself is a far cry from being a good way to specify visual elements and how they relate to each other.

I am loathe to make criticisms without also offering suggestions, so here goes. My own preference would be to use a client-side XSLT transform with an XHTML document to combine it with an SVG document containing appropriate placeholders, and thus render a compound SVG+XHTML document. All three of the above problems are eliminated in one fell swoop:

1. Because the presentation is completely specified in an independent XML format, one need no longer pollute one’s content markup with needless presentational elements.

2. XSLT is an extremely powerful selector language, providing possibilities which CSS cannot even dream of. (This also reduces the amount of presentational markup: you can easily specify, say, that the first ordered list of links encountered in a document is for navigation, without any recourse to IDs, classes or containers.)

3. SVG is swiftly becoming a well-respected graphics format in its own right. You want a drop-shadow for an irregular shape, or apply a transform matrix? SVG filters are more than up to the task.

Imagine the process. You start out with a nice, clean page of markup—no containers! Next, you open a graphics package and design your page layout, specifying which areas you want to stretch or scale and how, specifying a few placeholders where you want your content to go. Finally, with some help from clever software, you decide what rules should be used to decide which placeholder the content in your XHTML document should go into.

The technology already exists to do all this. XHTML, XSLT and SVG are fairly mature, well-defined standards, and implementation in non-IE browsers is pretty good. With a concerted effort, something like this could really work. One major concern is accessibility; for example, I’m not sure how a screenreader would cope with the compound rendering. Perhaps such software could be provided with the XHTML document itself (which, after all, is sent to the client side for transformation)? I just want to throw this out there and see what people think.

Why in the world would we ever dream of going back to that old-school nonsense? Furthermore, someone made an excellent point earlier – if MS wants to be taken seriously they should simply publish a patch that removes IE6 permanently, since it’s the root cause of this compatibility headache. Even this close to a gold release for IE8 I still find myself occasionally having to use conditional comments targeted at IE6, and CSS tables is not going to change any of that – it’s not a magic fix.

Lastly, why should those of us who embraced CSS 2.0 and standards want to re-learn a completely new layout standard when all it does it replicate something we’ve already discarded? To me the whole idea of CSS Tables is a non-issue; the only argument I’ve ever heard worth mentioning is that it’s tough to get matching columns with pure CSS. Shouldn’t we just fix this once in a new CSS spec and be done with it, instead of rethinking the whole layout structure to go back to a table environment? The whole approach just doesn’t make sense, I think it’s nuts.

I know well that not everybody can be a computer scientists and myself aren’t one of them. I’ve learned using CSS and other standard through tutorials because I feel myself more confident in programming languages than web standards. It was quiet difficult to understand the CSS standard itself as english is not my native. I think there is lack of good examples how to use the standards exactly. I could imagine good looking 3D movies explaining positioning and examples that points out the changes between the different attribute values. This is a kind of multimedia stuff . Even for developers of browers would be this a good reference. They can ask things like “Does the page appearing in my browser as in the movie?”. I’m quiet sure you finished discussing about similar things already.

best regards

That’s my opinion: English is a bad language to lay down any standards and definitions, it’s not exact enough

I definitely think that css tables are a better solution to design layout, and I will definitely be using them. But I will only implement them when there is enough browser support.
Until then, it’s floats and position for me…