[haberman]

The W3C is broken: it has spent the last 10 years standardizing ideas instead of existing practice. This is totally backwards and leads to standards that are too complicated, unrealistic, and in many cases not needed at all.

The W3C was originally created to standardize HTML, which was already being used by many vendors and users but in incompatible ways. That is exactly the right situation for creating a standard. It leads to standards that are realistic and motivated by a demonstrated need.

Unfortunately, almost everything else the W3C has ever done has happened in the opposite direction: in response to an idea or a perceived need, some people theorize about the best way to solve the problem and then write a document that a bunch of vendors are supposed to then implement from scratch.

This is how we ended up with the XML stack, which was designed to solve the data interchange problem but ended up being a disaster of complexity, inefficiency, and ad hoc implementations.

Even the case of CSS (which has been quite successful) is sub-optimal IMO, because it didn't choose to standardize the existing practice of how people were using tables for layout. The CSS box model makes it stupidly difficult to do things that are trivial with table-based layouts, like a a three column layout (which is considered a "holy grail" by even CSS advocates: http://www.alistapart.com/articles/holygrail/).

CSS could have used a table-like layout model that makes it easy to arrange <div>s into rows and columns, allowing a smooth upgrade from people who were using the <table> tag. Instead they invented something new that was much more difficult to design for, creating an unnecessary tension between web standards advocates and people who just wanted to get things done.

Standards should codify and refine existing practice, not attempt to invent new things.

[jamesbritt]

The W3C is broken: it has spent the last 10 years standardizing ideas instead of existing practice. This is totally backwards and leads to standards that are too complicated, unrealistic, and in many cases not needed at all.

The general view of the W3C as a standards body is broken.

The W3C gets companies and individuals together to hash out ideas for the Web. After some time and trial and error they assemble a recommendation. Hopefully people (i.e. the members of the W3C) put this into practice to see how well it flies in real-life. At some point, if the recommendation proves its worth, a standards body such as ISO or ANSI may want to formalize it. Or so it was.

Once upon a time yo could find on the W3C site a statement to the effect that they were not a standards body. However, some people who wanted to see the ideas from the W3C gain more traction (i.e. XHTML) started referring to W3C docs as standards, and the W3C as a standards body. Jeffrey Zeldman, in Designing with Web Standards [sic], says the use of the word "standards" when referring to these specs was, in fact, "a guerilla marketing maneuver."

This is not mere semantics. There is great value in making a distinction between well-considered ideas that still need to be battle-proven, and proper formal specs that have been sufficiently vetted.

Eventually the W3C began to believe the hype and now presents itself as a standards body, to the detriment of the greater Web community because, as you point out, they are "standardizing" things that people aren't actually doing.

[yuhong]

AFAIK, HTML has been standardized by the ISO based on the W3C HTML 4 recommendation.

[jamesbritt]

Indeed!

http://www.scss.tcd.ie/misc/15445/15445.HTML

[olavk]

I agree with you in general, but not in the case of CSS.

The CSS layout model were indeed based on exactly the layout properties of "presentational HTML" which people were using for style and layout at that time. It was just made more fine grained, and separated from the markup.

For example, the much hated "float/clear" layout model were based on the align="left|right" and clear="left|right" attributes which Netscape (I believe) introduce on img- and table-elements, and which people did use a lot. The layout model of tables were also widely used, and that was codified with the display:table-* properties, which again provides exactly the same layout model in CSS as tables provided via markup.

The problem was basically that IE development were frozen for many years. IE implemented the float-related CSS before the freeze, but not the table-model. I believe IE 8 were the first version of IE to support the table model, more than a decade after CSS2 became a recommendation.

This resulted in period of many years where the float model were supported via CSS in most browsers, but the table-model were not. This in turn resulted in an amazing amount of confusion, and people believing that floats are somehow the CSS's substitute for tables.

[flomo]

display:table- always felt more like a hack than as something part of the design.

As in "Oh, we forgot to model certain properties of tables that numerous websites actually use. Here's a workaround that imitates a table cell until we get around to figuring it out. PS: tables are still bad."

[olavk]

> PS: tables are still bad.

Tables are not bad.

CSS purists state that using table markup purely for layout purposes (e.g. for positioning a page footer) is bad, because it violates the separation between content and presentation. In that case display:table-* will give you the same layout (neither worse nor better!) but without requiring any specific markup in the html.

But this poses a dilemma for web developers, because display:table-* is not supported by all browsers, so some specifc kinds of layout can only be achieved with table markup.

Developers had to choose between (1) compromising the layout (2) compromising the semantic/style separation or (3) use complex workaround like the CSS frameworks, which tried to emulate the table-layout model without requiring display:table support in the browser. Neither is optimal.

[mattmanser]

I think you're misremembering history. It's not bad because 'it violates the separation between content and presentation', it's bad because you end up with tag soup. The table's demise was purely that it was too hard to maintain in anything more than a trivial layout. When you started getting down to your third nested table it was a friggin' nightmare. Also cells and rows had weird quirks as well as missing properties where you'd then have to nest a span in every cell just to get it behaving properly. Oh and also the occasional sacrifice to the dark god of 'table-layout:fixed'.

The rest of it though I agree with you.

[olavk]

But I think we agree. Layouts with nested tables are hard to maintain because presentational concerns (like the attributes which defines cell sizes) is spread all over the document and intermingled with the content. It's a mess.

[yuhong]

FYI, the reason why is that IE4 was released before CSS2 became a recommendation, and in fact implemented CSS1 (with bugs, but that is partly because it needs to be compatible with IE3) plus CSS-P: http://www.w3.org/TR/WD-positioning-19970819

[jerf]

"CSS could have used a table-like layout model that makes it easy to arrange <div>s into rows and columns, allowing a smooth upgrade from people who were using the <table> tag."

Like this? http://www.w3.org/TR/CSS2/tables.html

[haberman]

A lot like that, yes, but that was not introduced until CSS2 which is so big and complex that they've spent almost 15 years correcting it (CSS 2.1 keeps bouncing back and forth between "Working Draft" and "Candidate Recommendation").

Despite being first defined in CSS2 (1998) CSS tables were not supported in IE for 10 years (IE8, 2009), and while some of the blame for this is surely Microsoft's, with a standard as big and complex as CSS it's hard to expect implementations to get it right for a long time.

To avoid this disaster CSS3 is "modular" and currently has over 40 modules defined! This is madness. The more complicated things get, the more difficult it is to bring the web to new devices or to innovate in the browser space.

[halo]

CSS1 was deliberately designed to be 'simple' which is why they avoided the problem of layout entirely. Appendix E of the specification even says so (http://www.w3.org/TR/CSS1/#appendix-e).

[haberman]

Good to know -- though in that case, it seems inappropriate that web standards advocates were demanding everyone abandon table-based layouts when CSS did not actually support layout yet.

[bandushrew]

it was fairly inappropriate, and for the most part pragmatic web designers ignored it until it become realistic.

OTOH, it was the early adopters that were driving it, and I dont really have a problem with that. That is what early adopters do.

[tahu]

This time there is existing practice - Opera Unite and Tonido.

[yuhong]

Recently the W3C has recognized and fixed the problem with a requirement that two implementations has to interoperably implement a feature for it to survive the Proposed Recommendation. That is how CSS 2.1 got created after CSS 2.0 that became a recommendation in 1998 ended up with no browsers fully complying with it.

[alanh]

Have you ever heard of HTML5? It is fundamentally a codification & standardization of existing practice (<canvas>, parsing, short encoding meta tags) — and when it isn’t doing that, it addresses actual needs (like how to handle video and audio).

[haberman]

HTML5 was created by WHATWG (not W3C) in 2004. WHATWG was formed specifically because W3C had abandoned HTML in favor of XHTML. Only when it was clear that XHTML was a failure did W3C abandon it and join forces with WHATWG in 2007.

In other words, W3C missed the boat on this. I don't think it's to their great credit that they joined a three-year-old effort after it was already clear that the whole industry was headed towards HTML5 and not XHTML.

[thaumaturgy]

Are you seriously asking someone on HN if they've ever heard of HTML5?

[alanh]

Rhetorical, and if I could still edit to take the tone a step towards neutral, I would. But parent is an out-of-place and out-of-date criticism of w3c in my opinion. At minimum, they should explain why HTML5 is not a counter-example of their claim

[epistasis]

You should learn the history of HTML5, it's a perfect example of how the W3C is terribly broken, and certainly not an adequate or passable standards body. HTML5 was developed outside of W3C, in spite of the W3C, and at the objection of the W3C. Remember XHTML2? Hopefully not, but that's what the W3C wanted to do. The WHATWG formed to create HTML5 because browser makers wanted to improve the web and it was obvious that XHTML2 wasn't going to do that.

After it was clear how useful HTML5 was, the W3C was dragged into participation, but unlike W3C standards, HTML5 continued on with it's open development model where discussion takes place in the open.

The W3C has not contributed the idea, the execution, the prolitical process, or the people to HTML5, and HTML5 is a W3C standard by label only. If you look at the other standards the W3C creates, you'll see a pattern where standards are developed in isolation behind closed doors and without implementations, in direct contrast to any useable and useful standard such as HTML5. RDF, OWL, SVG, all abominations that seem to exist to hold back others from making reasonable and useful replacements.

[rimantas]

CSS offers many solutions to layout. The problem is what a lot of web surfers are stuck with old versions of some particular browser.

[haberman]

What solution does CSS offer for three-column layout (with a resizable center column) that is remotely as simple as:

    <table width=100%><tr>
      <td width="200px">Fixed left column.
      <td>Liquid center column
      <td width="200px">Fixed right column.
    </table>

[drdaeman]

Upcoming CSS3 features (some browsers support some, but not others).

    <section id="box">
        <div class="left">...</div>
        <div class="center">...</div>
        <div class="right">...</div>
    </section>

- Template layout module (W3C)

    #box { display: '  a   b  c'
                     200px * 200px;
           width: 100%; }
    #box .left { position: a; }
    #box .center { position: b; }
    #box .right { position: c; }

- Flexible box model (Mozilla)

    #box { display: box; box-orient: horizontal; width: 100%; }
    #box .left, #box .right { width: 200px; }
    #box .center { box-flex: 1; }

- Grid positioning module (Microsoft)

    #box { columns: 3; grid-columns: 200px * 200px; }
    #box.left, #box.right { width: 200px; }

Disclaimer: I didn't test anything, just wrote the code from memory, peeking at random on-subject Google results. It is possible that I forgot, misspelled or misunderstood something.

[kleiba]

That is because you picked an example where an old school HTML table shines and CSS looks complicated. But likewise there are many other things you can do with CSS that are pretty hard or impossible to do without it.

If webpages were always only made of box-like structures, CSS would have never made it. But you will admit that it offers a web designer much more flexibility.

That flexibility comes at a cost though, because some examples (like yours) that were very easy to realize with the very limited layout options pure HTML offered have now become more difficult. But that's because it provides a more general approach that takes care of many problems, not just three-column layout, and must thus be flexible. And flexible is what old timer HTML was not.

I'm not saying that CSS is the best or even a very good solution to the problem. I'm sure there are a lot of different approaches how to address the general problem of dividing content and layout. But it's there now, and it's bearable, I think.

Besides... all of what I wrote has nothing to do with the w3c :-)

[haberman]

I looked into CSS tables a bit more, and this seems to be how you'd write the above:

    <style type="text/css">
    .table { display: table; width: 100%; }
    .tr    { display: table-row; }
    .left-column { display: table-cell; width: 200px; }
    .middle-column { display: table-cell; }
    .right-column { display: table-cell; width: 200px; }
    </style>
    <div class="table"><div class="tr">
      <div class="left-column">Fixed left column.</div>
      <div class="middle column">Liquid center column.</div>
      <div class="right-column">Fixed right column.</div>
    </div></div>

This is definitely a huge improvement over the CSS box model. If this had been available in practice 10 years ago, I probably wouldn't be complaining about it now. But I still think it is because CSS2 is so big, complicated, and had so many errors that it wasn't implemented for so long.

[BrandonM]

That's so much more verbose than using tables. What do you gain?

[olavk]

Regarding verbosity: In any real wold page, you would have some additional css properties, so the overhead with defining classes and referencing them in css would be there anyway. So the verbosity advantage of html tables is down to "tr" being shorter that "display: table-row". That is not really a serious objection.

Btw., if you don't want a seperate stylesheet, you can embed the css style information directly in the html:

    <div style="display: table; width: 100%"><div style="display: table-row;">
      <div style="display: table-cell; width: 200px;">Fixed left column.</div>
      <div style="display: table-cell;">Liquid center column.</div>
      <div class="display: table-cell; width: 200px;">Fixed right column.</div>
    </div></div>

This quickly becomes an unmaintainable mess but if you really want, you can do it.

[olavk]

It gives you separation of concerns between content structure and presentation. This allows you to adapt the presentation for different audiences. For example, on a phone with a small screen, you might not want a three-column layout, but just have everything in one column. You can do that with an alternative style-sheet.

Additionally, you don't have to confuse screen readers with a table which don't really contain tabular data. So it also more accessible.

[mdaniel]

Without actually trying that code, my guess is that unlike a formal "table", those 3 div tags will have different page-wrap behavior (e.g. not intra-div, but the wrapping of the div elements with respect to the screen width).

That may not appear to be a big deal, but try loading that page on a mobile device and it will become a bigger deal.

[null_ptr]

    <div id="site">
        <div id="left">Fixed left column.</div>
        <div id="middle">Liquid center column</div>
        <div id="right">Fixed right column.</div>
    </div>

The whole point of CSS is to take things like width="200px" out of the markup.

[nokcha]

>The whole point of CSS is to take things like width="200px" out of the markup.

I don't see how that's really an improvement in this particular case. In fact, I would say that using 'width="200px"' inline may actually be more readily understandable and maintainable than having the style defined somewhere else and then used only once. (I'm assuming that the 'width="200px"' is not duplicated multiple times in the PHP/Python/Arc/etc code that generates the HTML page. If the same style needs to be identified multiple times in the source code, then of course it may be a better idea to define it once and thereafter refer to it by name.) Are there any practical benefits that I may be missing?

[billybob]

How do you know when you use a style the first time that you'll never use it elsewhere? If you've got many one-off style declarations, you probably don't have a very consistent graphical style.

Besides that, it's nice to know where all your styles live instead of having to grep your whole project. And inline styles take precedence over everything else, so mixing them with stylesheets can create havoc.

[olavk]

The example is contrived because there is no content in the table cells. In the real world there will be some content, and suddenly your "width"-attributes are many lines apart and not nicely aligned. This makes is a mess to maintain, because there is no single place where your layout is defined; the information is spread all over the document, mixed with other kinds of information.

Furthermore, separating style info into a style sheet allows you to adapt the style for different devices.

[haberman]

You have to show what your CSS would be also. My point is that the CSS you have to write to get your example to look like mine is extremely complicated when all you have is the box model.

[Groxx]

Like this:

  #left{
    float:left;
    width:200px;
    margin-right: -200px;
  }
  #right{
    float:right;
    width:200px;
    margin-left: -200px;
  }
  #middle{
    float: left;
    margin-left: 200px;
    margin-right: 200px;
  }

People tend to avoid this relatively simple solution because of the double-left-margin bug in IE 5 & 6, but that's getting to be an excessively old set of browsers. And even then, it's usually fixed with a "display:inline;" on the ones with left margins.

edit: it's not an ideal solution, as it has those negative values which have to match the related positive ones. Which is solved with something like Sass or Less, because CSS lacks any "programming" abilities, which I see as its main failure. Even variables and simple math would be acceptable, but instead, nothing.

[Groxx]

Downvote? It's correct, and it's exactly what was asked for. Here's a Fiddle demonstrating it: http://jsfiddle.net/yBs5V/

[kerryfalk]

Or, to make it _really_ simple:

  <div class="column">1</div>
  <div class="column">2</div>
  <div class="column">3</div>

  div.column {
    margin: 10px;
    width: 200px;
    float: left;
    }

EDIT: Missed the key constraint, resizable center column. What I have above won't work for that, but shows that replacing tables isn't very difficult.

[rimantas]

There is nothing complicated, even using floats. Usig display: table-* it's damn easy. You also got CSS columns, and flex box.

[flomo]

> The whole point of CSS is to take things like width="200px" out of the markup.

IIRC, you could stick the TD's width value in a CSS class and it would work exactly the same.