Semantic HTML isn't only about the elements we use. It's quite obvious that we should use <a> for links, <table> for tabular data and <p> for paragraphs etc. What's less obvious is the names we use for classes.

As Phil Karton says, “there are only two hard things in Computer Science: cache invalidation and naming things.” So spending an entire chapter talking about it seems like an appropriate thing to do.

Naming is quite frankly the most important aspect of writing maintainable CSS. There are two main approaches: the semantic approach and the non-semantic approach. Let's discuss what they are.

Semantic vs non-semantic

<!-- non semantic -->
<div class="red pull-left pb3">
<div class="grid row">
<div class="col-xs-4">

<!-- semantic -->
<div class="basket">
<div class="product">
<div class="searchResults">

Non-semantic classes don't convey what an element represents. At best, they give us an idea of what an element looks like. Atomic, visual, behavioural and utility classes are all forms of non-semantic classes.

Semantic classes don't convey their styles, but that's okay. That's what CSS is for. Semantic classes mean something to HTML, CSS, Javascript and automated functional tests.

Reasons why semantic classes work better:

1. Because they are readable

Here's a real snippet of HTML using atomic classes:

<div class="pb3 pb4-ns pt4 pt5-ns mt4 black-70 fl-l w-50-l">
  <h1 class="f4 fw6 f1-ns lh-title measure mt0">Heading</h1>
  <p class="f5 f4-ns fw4 b measure dib-m lh-copy">Tagline</p>

Here's the same thing using semantic classes:

<div class="hero">
  <h1 class="hero-title">Heading</h1>
  <p class="hero-tagline">Tagline</p>

2. Because it's easier to build responsive sites

Imagine coding a two-column responsive grid whereby:

With visual/utility classes it looks like this:

<div class="grid clearfix">
  <div class="col pd20 pd50 fs2 fs3">Column 1</div>
  <div class="col pd20 pd50 fs2 fs3">Column 2</div>

A quick evaluation shows significant pain already.

With semantic classes it looks like this:

<div class="thing">
  <div class="thing-thingA"></div>
  <div class="thing-thingB"></div>

Question: How valuable is a codified responsive grid system? A layout should adapt to the content, not the other way around.

3. Because they are easier to find

Searching for HTML with a non-semantic class yields many results. As semantic classes are unique, a search yields only one result, making it easy to track down the HTML.

4. Because they eliminate the risk of regression

Updating a visual class could cause regression across a multitude of elements. Updating a semantic class only applies to the module in question, eliminating regression altogether.

5. Because visual classes aren't worth it

In some respects we may as well inline styles. This is more explicit and reduces the CSS footprint to zero. Inline CSS is a problem though, because we can't use media queries for example. And placing CSS in HTML mixes concerns and removes the ability to cache it.

Question: Isn't .red the exact same abstraction that CSS already gives us for free with color: red?

6. Because they provide hooks for automated tests

Automated functional tests work by searching for, and interacting with elements. This may include:

  1. clicking a link
  2. finding a text box
  3. typing in text
  4. submitting a form
  5. verifying some criteria

We can't use non-semantic classes to target specific elements. And adding hooks specifically for tests is wasteful as the user has to download this stuff.

7. Because they provide hooks for Javascript

We can't use non-semantic classes to target specific elements in order to enhance them with Javascript.

8. Because they don't need maintaining

If we name a thing based on what it is, we won't have to update the HTML again e.g. a heading is always a heading, no matter what it looks like.

With visual classes, both the HTML and the CSS need updating (assuming there aren't any selectors available for use).

9. Because they are easier to debug

Inspecting an element with a multitude of atomic classes, means wading through many selectors. With a semantic class, there's only one, making it far easier to work with.

10. Because the standards recommend it

On using the class attribute, HTML5 specs say in

"[...] authors are encouraged to use values that describe the nature of the content, rather than values that describe the desired presentation of the content."

11. Because styling state is easier

Consider the following HTML:

<a class="padding-left-20 red" href="#"></a>

Changing the padding and colour on hover is a difficult task. It's better to avoid having to fix self-induced problems like this.

12. Because they produce a small HTML footprint

As we've seen above, atomic classes bloat HTML. Semantic classes result in smaller HTML. And whilst the CSS may increase in size, it's cacheable.

Final thought

Semantic classes are a corner stone of MaintainableCSS. Without them, everything else makes little sense. Name something based on what it is and everything else falls into place.


  1. Introduction
  2. Semantics
  3. Reuse
  4. IDs
  5. Conventions
  6. Modules
  7. State
  8. Modifiers
  9. Versioning
  10. Javascript
  11. Organisation
  12. FAQs

Want updates?

Join my newsletter to get the latest MaintainableCSS updates and my usual articles about design and interface development.