Why not use tables for layout in HTML?

Obvious answer: See CSS Zen Garden. If you tell me that you can easily do the same with a table-based layout (remember - the HTML isn't changing) then by all means use tables for layout.

Two other important things are accessibility and SEO.

Both care about in what order information is presented. You cannot easily present your navigation at the top of the page if your table-based layout puts it in the 3rd cell of the 2nd row of the 2nd nested table on the page.

So your answers are maintainability, accessibility and SEO.

Don't be lazy. Do things the right and proper way even if they are a bit harder to learn.

I'm going to go through your arguments one after another and try to show the errors in them.

It's good to separate content from layout But this is a fallacious argument; Cliché Thinking.

It's not fallacious at all because HTML was designed intentionally. Misuse of an element might not be completely out of question (after all, new idioms have developed in other languages, as well) but possible negative implications have to be counterbalanced. Additionally, even if there were no arguments against misusing the <table> element today, there might be tomorrow because of the way browser vendors apply special treatment to the element. After all, they know that “<table> elements are for tabular data only” and might use this fact to improve the rendering engine, in the process subtly changing how <table>s behave, and thus breaking cases where it was previously misused.

So what? Does my boss care? Do my users care?

Depends. Is your boss pointy-haired? Then he might not care. If she's competent, then she will care, because the users will.

Perhaps me or my fellow developers who have to maintain a web page care... Is a table less maintainable? I think using a table is easier than using divs and css.

The majority of professional web developers seem to oppose you^{[citation needed]}. That tables are in fact less maintainable should be obvious. Using tables for layout means that changing the corporate layout will in fact mean changing every single page. This can be very expensive. On the other hand, judicious use of semantically meaningful HTML combined with CSS might confine such changes to the CSS and the pictures used.

By the way... why is using a div or a span good separation of content from layout and a table not? Getting a good layout with only divs often requires a lot of nested divs.

Deeply nested <div>s are an anti-pattern just as table layouts. Good web designers don't need many of them. On the other hand, even such deep-nested divs don't have many of the problems of table layouts. In fact, they can even contribute to a semantic structure by logically dividing the content in parts.

Readability of the code I think it's the other way around. Most people understand html, little understand css. It's simpler.

“Most people” don't matter. Professionals matter. For professionals, table layouts create many more problems than HTML + CSS. This is like saying I shouldn't use GVim or Emacs because Notepad is simpler for most people. Or that I shouldn't use LaTeX because MS Word is simpler for most people.

It's better for SEO not to use tables

I don't know if this is true and wouldn't use this as an argument but it would be logical. Search engines search for relevant data. While tabular data could of course be relevant, it's rarely what users search for. Users search for terms used in the page title or similarly prominent positions. It would therefore be logical to exclude tabular content from filtering and thus cutting the processing time (and costs!) by a large factor.

Tables are slower. An extra tbody element has to be inserted. This is peanuts for modern web browsers.

The extra element has got nothing to do with tables being slower. On the other hand, the layout algorithm for tables is much harder, the browser often has to wait for the whole table to load before it can begin to layout the content. Additionally, caching of the layout won't work (CSS can easily be cached). All this has been mentioned before.

Show me some benchmarks where the use of a table significantly slows down a page.

Unfortunately, I don't have any benchmark data. I would be interested in it myself because it's right that this argument lacks a certain scientific rigour.

Most web sites that need an upgrade need new content (html) as well. Scenarios where a new version of a web site only needs a new css file are not very likely.

Not at all. I've worked on several cases where changing the design was simplified by a separation of content and design. It's often still necessary to change some HTML code but the changes will always be much more confined. Additionally, design changes must on occasion be made dynamically. Consider template engines such as the one used by the WordPress blogging system. Table layouts would literally kill this system. I've worked on a similar case for a commercial software. Being able to change the design without changing the HTML code was one of the business requirements.

Another thing. Table layout makes automated parsing of websites (screen scraping) much harder. This might sound trivial because, after all, who does it? I was surprised myself. Screen scraping can help a lot if the service in question doesn't offer a WebService alternative to access its data. I'm working in bioinformatics where this is a sad reality. Modern web techniques and WebServices have not reached most developers and often, screen scraping is the only way to automate the process of getting data. No wonder that many biologists still perform such tasks manually. For thousands of data sets.

Here's my programmer's answer from a simliar thread

Semantics 101

First take a look at this code and think about what's wrong here...

class car {
    int wheels = 4;
    string engine;
}

car mybike = new car();
mybike.wheels = 2;
mybike.engine = null;

The problem, of course, is that a bike is not a car. The car class is an inappropriate class for the bike instance. The code is error-free, but is semantically incorrect. It reflects poorly on the programmer.

Semantics 102

Now apply this to document markup. If your document needs to present tabular data, then the appropriate tag would be <table>. If you place navigation into a table however, then you're misusing the intended purpose of the <table> element. In the second case, you're not presenting tabular data -- you're (mis)using the <table> element to achieve a presentational goal.

Conclusion

Will visitors notice? No. Does your boss care? Maybe. Do we sometimes cut corners as programmers? Sure. But should we? No. Who benefits if you use semantic markup? You -- and your professional reputation. Now go and do the right thing.