I have been told, as well as read that using the style attribute in html is considered bad/sloppy/poor form. Further, that all rendering specific bits should be divorced into css and other parts as appropriate. I am trying to understand why exactly this is.
I can see why you might want to keep the HTML a pure semantic DOM, that speaks about the structure of the document, but for practical pages, the importance is that the page looks right and functions appropriately.
Is there some more compelling reasons for this separation?
Separation of concerns This makes it easy to replace the styles without changing the markup, or vice versa. Plus you can have one person working on CSS and another working on content
Don't Repeat Yourself You can apply a style to many elements without having to repeat it over and over. Smaller pages means quicker load times using less bandwidth. Plus it's easier to modify later, because you can change it in one place in one file, instead of many places in many files.
Cachability If the same style sheet is used on every page of your site, browsers can download it once and then cache it, instead of downloading the styles with the content of every single page. And they won't have to re-download it whenever the content of those pages changes.
Multiple Versions It is easy to create multiple versions of the visual layout and appearance of your site since you just need to swap out the stylesheet file to change the appearance of every page. For instance, you can create a white-label version of a web application which your partners can re-skin to match their brand. See CSS Zen Garden for some great examples of how flexible this approach can be.
Start with this code:
<ul>
<li style="color: blue;">One</li>
<li style="color: blue;">Two</li>
<li style="color: blue;">Three</li>
<li style="color: blue;">Four</li>
</ul>
Let's say that today, you decide to change the link color to red. Since these styles are inline, you tediously have to walk through each element and change the style attribute. Imagine doing this for 10, maybe 20 HTML pages and you'll see why this becomes a problem.
Using a stylesheet separates the content:
<ul>
<li>One</li>
<li>Two</li>
<li>Three</li>
<li>Four</li>
</ul>
From the style:
ul li {
color: blue;
}
Had you used a stylesheet from the beginning, changing the color is as simple as changing blue to red in your stylesheet.
Aside from making the document easier to style, there's also selector specificity. Imagine that you inherited the first code chunk from a previous developer and would like to change the color again, but you (being a nice developer) prefer stylesheets:
ul li {
color: red;
}
You'll soon become frustrated and resort to using !important, as your selectors can't override the inline styles.
CSS should be another file included in HTML because, if you want to change one style of an element that is included in more than one pages you will just change one style from CSS and the changes will be applied to all of the files. If you have the style in HTML, you would need to go on the pages one by one and change the styling. Its a good template building practice.
By separating markup and css. You can use css to change the look of everything, without affecting the markup.
Benefits include:
Creating different designs for the same html.
Dividing work within a team. One front-end developer can focus entirely on the css.
Back-end developers, do not have to hassle with the css.
Easier to change the look in the future.
Easier to migrate the html-markup to a new platform or content management system in the future.
Lots of websites use class names like floatleft, clearfloat, alignright, small, center etc that describe the style that is attached to the class. This seems to make sense so when writing new content you can easily wrap (for example) <div class="clearfloat">...</div> around your element to make it behave the way you want.
My question is, doesn't this style of naming classes go against the idea of separating content from presentation? Putting class="floatleft" on an element is clearly putting presentation information into the HTML document.
Should class names like this that directly describe the attached style be avoided, and if so what alternative is there?
To clarify, this isn't just a question of what to name classes. For example a semantically accurate document might look something like:
<div class="foo">Some info about foo</div>
...
<div class="bar">Info about unrelated topic bar</div>
...
<div class="foobar">Another unrelated topic</div>
Say all these divs need to clear floats, the css would look something like:
div.foo, div.bar, div.foobar {
clear:both;
}
This starts to get ugly as the number of these clearing elements increases - whereas a single class="clearfloat" would serve the same purpose. Is it recommended to group elements based on the attached styles to avoid repetition in the CSS, even if this means presentational information creeps into the HTML?
Update: Thanks for all the answers. The general consensus seems to be to avoid these class names in favour of semantic names, or at least use them sparingly provided they don't hinder maintenance. I think the important thing is that changes in the layout should not require excessive changes to the markup (although a few people said minor changes are okay if it makes overall maintenance easier). Thanks to those who suggested other methods to keep CSS code smaller as well.
It's great until you re-design, and narrow is highlighted yellow, center converts better left-justified, and the image you called floatleft now belongs on the right.
I'll admit to the sin of using floatleft and clear as CSS class names, but it is much easier to maintain your CSS if you choose names that relate to the semantic meaning of the content, like feedback and heroimage.
Presentational class names
The HTML spec is clear on this issue:
There are no additional restrictions
on the tokens authors can use in the
class attribute, but authors are
encouraged to use values that describe
the nature of the content, rather than
values that describe the desired
presentation of the content.
Does clearleft describe the nature of the content? Not really. Eric Meyer made a joke about this a while ago.
Try to find a structural relation between the seemingly unrelated elements
Let's say you have paragraphs about ducks, paragraphs about monkeys and paragraphs about frogs. You want them have a blue background.
<p class="duck"></p>
<p class="monkey"></p>
<p class="frog"></p>
You could add this CSS rule:
p.duck, p.monkey, p.frog {
background-color: blue;
}
But aren't they all animals? Just add another animal token:
<p class="animal duck"></p>
<p class="animal monkey"></p>
<p class="animal frog"></p>
And change the CSS rule to:
p.animal {
background-color: blue;
}
It is hard and it might not always be possible but the important thing is not to give up quickly.
What if you can't?
If you have a lot of elements with absolutely no structural relation between them, that indicates a structural problem with your document. Try to decrease these elements. That said, grouping n CSS selectors on one rule is still better than adding n presentational class tokens in your HTML document.
Style classes should be semantic. This is a great article on semantic web page design (well, I found it really helpful anyway).
EDIT: I just read another article that makes some good points for using things like display: inline-block, display: table etc. instead of floats. That should help avoid those pesky floatleft and clearfix classes. Making them semantic is always up to you though.
The main problem with having classes named floatleft, clear or the like is the fact that changes in the design imply changes in the HTML markup. This should not happen, true separation between content and presentation is achieved only when you can re-use the same markup in multiple designs or even media (think sharing the same HTML between desktop and mobile versions of your site and only switching sylesheets).
Now, for a practical example :). To add on Fredrik's answer, LESSCSS allow you to hide styles declarations/mixins from developers. This way you can protect reusable components in your stylesheets without the danger of having them pop up in your HTML.
Sample markup:
<div class="branding">Company Name</div>
Sample less code:
// example of reusable functions
.noText() {
color: transparent;
text-indent: -9999px;
}
.clear-after() {
&:after {
content: ".";
display: block;
visibility: hidden;
height: 0;
clear: both;
}
}
.branding {
.clear-after();
.noText();
background-image: ...;
}
Of course, for mobile you might just want to have the company name in bold, without any other styling:
.branding {
color: pink;
font-weight: bold;
}
I think it depends on how you are using the styles.
Content should be named accordingly, as the style may change but the content will likely remain the same.
For instance, if you have a div that contains stock info, you should name the div something like div class="stockInfo", so that no matter what the presentation, you can change the styles and the name will not contradict those styles (as opposed to naming the div div class="yellow" and then changing the background-color to red).
However you will have "helper styles" and these should be named for what they do.
For instance, you will likely want to use a <br /> to clear some floats. In this case, it is perfectly reasonable to name it <br class="clear" /> and to give it a style of br {clear:both;}.
Again, most Web sites float their images right or left. To assist with this, you can set <img class="right" src="" /> and <img class="left" src="" /> and then have the styles to match, img.right {float:right;} etc.
So it depends on the usage.
Class names and ids that describe the function is better than using names that describe the styling of the element.
I usually end up not doing it religiously though, because it is in my opinion more convenient to i.e. clear floating elements by using the famous clearfix hack rather than adding clear:both all over the stylesheets.
But I think that LESS and SASS creates interesting opportunities to get the best out of both worlds, because you can have mixins and functions that describes some style and still have semantic correct names by just including whatever 'style' you want.
Instead of having this HTML and CSS
<div class="navigation roundedcorners">...</div>
.roundedcorners {
-moz-border-radius: 5px;
-webkit-border-radius: 5px;
border-radius: 5px;
}
you could use SASS to create this mixin:
=rounded-corners
-moz-border-radius: 5px
-webkit-border-radius: 5px
border-radius: 5px
and include it into your .navigation class like this:
.navigation
+rounded-corners-5px
which would reduce your HTML to this:
<div class="navigation">...</div>
and therefore still get the advantage of having semantic correct names while having a convenient way to share styles between different elements.
Andrew; it's good to give sensible name to an class & id which easy to understand for you & your fellow member's which are working on that project. For me classes small , center , floatleft etc define nothing to me because when you give class center that's indicate that the element on the center but there are other properties also in that class like color, background etc
For example
<div class="wrap">
<div class="center">lorem</div>
</div>
css:
.center{margin:0 auto;}
in this example class center don't clear to me. but we can use them as a helper class.
For example
<div class="wrap">
<div class="panel center narrow">lorem</div>
</div>
css:
.center{margin:0 auto;}
from above example now it clear to me what the role of class center in that panel div
FOR MORE CHECK THESE LINKS :
What's the best way to name IDs and classes in CSS and HTML?
http://www.ronniesan.com/blog/entry.php?title=organizing-your-dom-elements-with-the-proper-ids
http://cssglobe.com/post/3745/my-top-10-most-used-css-class-names
If your question is:
Is it recommended to group elements
based on the attached styles to avoid
repetition in the CSS, even if this
means presentational information
creeps into the HTML?
Then my flat answer would be that in real world, semantic and presentation isn't everything. So my answer would be: it depends.
...depends if bandwidth is important to you... On a site with many visitors per hour, then the classnames could even simply be something like "c11" (yes, I've seen it) instead of meaningful, but looong class names.
...depends also if you perfectly know that when the look and feel will change, then the CODE will change also. (exemple: you redesign a site today in XHTML, but you perfectly know that when you'll re-do the CSS in 2 years, then you'll want the markup to be HTML5, so you will pretty much change the structure anyways...)
...depends also if you're already 3 days late on a delivery. When you're 3 days late, believe me, classnames like "nopadding" starts to appear, as you don't have anymore time to think straight about semantic (neither does your client).
Depends on so many things, I'd say... That's my "real life" point of view of your question.
From what I've seen, developers have the tendency to overload their HTML pages with way too many unnecessary classes and extra markup. These classes not only make the page size bigger (and thus the load time longer), they also crowd up the page and make it difficult to manage and modify at a later time.
Having things like center and float-left might be helpful when you're dealing with display text that was input by a user (such as a post on a forum), but for general markup purposes you're better off just adding text-align: center and float: left to the appropriate classes. This especially helps if you are trying to change the appearance of your site without changing the HTML much. The less you have hardcoded into your template, the easier it is to only have to change around the CSS when modifying your template. That bit alone is worth it to me.
As a general rule of thumb, you really should only give elements classes when it describes what the content is, not where or how it is being displayed. i.e. <span class="movie-information"> instead of <span class="bold">.
The only time I feel that it makes sense to give an element a class when it isn't necessary is if you are concerned with Search Engine Optimization. You should definitely read up on Microformats if you are interested in seeing how adding the right classes can actually help search engines. That being said, adding classes that describe how the page is visually displayed does nothing for the search engines.
The only time I would ever "group" classes is if they are either displaying the same thing, or if they are siblings. It gets really messy in your CSS when you have elements from all over your page defined together. You're much better off grouping your classes in your stylesheet in ways that you will be able to find them later, rather than saving a few lines by combining them.
I think this is where old meets new in web technologies. From times past, it has been difficult to unobtrusively render an outstanding web experience. These class names mostly came in handy when websites were changing webmasters to aid them in understanding the code. It served its cause well but with the new technologies of this day and age, I think this is slowly dying out - infact, it should be dead.
The question we should ask is, "Do we need to create a new class for every new innovative design that could pass as a template?". I do not think so. The markup on a site should do what it is meant for - markup. The class names used in the markup should be descriptive of the content and not its looks. The stylesheets - on the other hand - should be able to select elements on a document based on the information in the markup, and style them.
I want to relate this to the Rails association naming convention. Consider this...
class Post < ActiveRecord::Base
has_one :personifyable
has_many :personifications, :through => :personifyable
has_many :taggables
has_many :tags, :through => taggables
belongs_to :destroyers
end
Obviously, this is not a real model; it is something I am using to drive a point. Consider the case of a deeply nested dependency. These names will grow ridiculous - if they aren't already (i.e. in CSS, <div class='mediumwidth floatright centeredtext graytheme master'></div> or something of the like)
Now consider the case where you have different principles. Different developers and designers may - if not 'most definitely will' - have different reasons for using a specific naming convention. How would this affect refactoring time. I will leave that to your imagination. Also, if your business partner notices a new trend with sites themes that attracts traffic - more technically, assume this business partner has performed some experimental A/B testing and come up with some specs - you don't want to change the contents of the whole stack (ie HTML and CSS and possibly JS pages) to implement this new style.
In conclusion, keep styling hints out of your markup. Unobtrusively interact with the document to style and manipulate it. Sass gives you a fine way of styling a site while having your CSS mock your markup. jQuery is another awesome UJS library. HTML5 gives you methods too that make the markup more flexible and yields more information to CSS and JS.
I don't think adding a descriptive class name to your document is really a big issue. I find it's easier to work with explicit class names like "floatleft" as opposed to things that are purely semantic or reliant on the cascade. It's usually easier for later developers who don't have the document structure in their heads as well.
You don't want to use them for everything--you wouldn't want to add a class of floatleft to every li in a left-floated menu, but these kind of styles are very good when you need to do a specific thing to one or more elements, and you want to make other developers aware that you did it.
It's like putting in <div class="clear"> or even <div style="clear:both;">: maybe not the prettiest but it sure is obvious what you are doing.
My rule of thumb is: whatever makes you have to think less, do that.
EDIT: As I said in my comment above, this is most true for classes that refer to clearing and floats, i.e., things that are purely presentational, non-semantic, and yet must be referred to in the HTML. I think in this case it is actually preferable to indicate that you are using a purely presentational class, like floatleft, rather than forcing the float to be attached to some semantic element.
I am a developer before a programmer, so for me I use something like a "floatleft" css class as a sort of UtilityMethod.
Meaning, my css class is "floatleft"...and that's all that class does.
so if I say <div class="floatleft"></div> in my mind that is saying "make this div float to the left".
So if that Div also needs a blue background and is my main header it's going to have a different class for that and I end up with <div class="mainheader floatleft"></div>
Doing it this way also avoids issues with refactoring. If I am changing my design later, I will know that "floatleft" ALWAYS floats things left and nothing more.
I think, at the end of the day it is about what works for you. If your class name is descriptive of what it does, that does not really go against the rule of separating markup from styles. Another factor to consider is, are you the sole developer, or part of a team. If you are part of a team, or you know your code will be worked on later by other devs, you should establish, and document the naming conventions used.
I am currently contracting with Down Jones on some very large projects, and we have a rather lengthy document on naming conventions for our classes, including when to use camel-case, or dashes or underscores, as well as specific prefixes on the class name based on the project we work on. It sounds crazy, but when you have a dozen or so front-end devs working on stuff at the same time, it is a life saver!
You are saying something like this:
.red
{
color:red;
}
so in order to use this class:
<ul>
<li class="red">hello</li>
</li>
ALTERNATIVE SOLUTION
ul li
{
color:red;
}
Usage:
<ul>
<li>Hello</li>
</ul>
By this you can actually remove the presentation information from the content.
I personally name them things close to what they will be doing. Say I have a class that is on an image gallery and its a primary most used class it will be something like "gallery" or if I'm setting borders around things that are meant to be more decorative I'll name it "decoborder". I try to keep them semi short and somewhat related what task they provide. I don't like to do things like "small, large, H1underlined" or anything that can mimick another tag or function because that can just get confusing. Beyond that I think you should really name it in whatever way makes the most sense to you.
If the question is only one of naming, then for one specific class...
class="floatleft"
or
class="myClass"
or
class="gibberish"
....changes absolutely nothing. They are only different class names. The programming functions the same.
Either your content & presentation is separated, or it isn't... totally regardless of how you created the names.
For what it's worth, if I remember correctly the class keyword in HTML isn't currently used for anything other than CSS stylesheets. Thus, the example that you provided...
<div class="foo">Some info about foo</div>
...
<div class="bar">Info about unrelated topic bar</div>
...wouldn't really be a way of identifying data. I would suggest either the name or id attribute if you really want to identify data within your HTML tags. (They both have slightly different uses - name is generally used for server-side queries while id can be styled and is generally more versatile. IDs must be unique, but names don't have to be.) You can check further documentation using the W3C HTML specification.
In short - don't worry about tying content to presentation through your tag classes; until they're specifically used for anything else, they will not have any real effect on your raw content. As such, I'd say name your classes whatever you want, as long as it makes sense. Personally, I would err on the side of logical naming versus style-type naming (e.g. class name "editorcomment" instead of class "graybgfloatleft" or something like that), but in the end, your class names aren't going to tie your data to your presentation like an ID or a name would.
Good luck.
I've done both and I have to say nowdays I lean towards using non presentational classnames. I found this nifty framework called OOCSS https://github.com/stubbornella/oocss/wiki which helped me alot when I was creating a new layout for my web application and suited my requirements so well.
It is mostly because the definition of basic classes for spacing, headers and text works so well when you have to deal with alot of content. Because you use the same classes everywhere it helps make the layout better and maintainable.
Offcourse this means a element in your html can look like this: <div class="unit size1of3 lastUnit"> But isnt that what HTML is about? Making sure your information is presented properly.
I disagree on the whole redesign point, be honest, when you have to redesign your website most CSS goes out the door anyways. And, by dividing the CSS into proper classes for spacing/heading/text it becomes less likely to have conflicting css rules that mess stuff up like ul li div{}
Offcourse the classes do not describe the contents, but as CSS does not allow class inheritance and you have to support old technology like IE6...does it really matter? And do classnames like animal and duck really make for better html? Id like to think HTML is for the browser and when the browser renders it, thats for humans.
It depends, sometimes it makes sense just to add a class to let an element float. The problem with the semantic approach is that you will end up ball of mud of css classes. Sure, names like redLink or blackHeader have to be banned but sometimes you will need little helpers like "clear" or "floatLeft".
Read this article of Nicole Sullivan who explain this in deep.
There are two things that I feel get entirely left out of these discussions all too often. The first is WHY you would want to be all semantic or all not. The keywords are Branding and Skinning. Presentational class names may be justifiable if you work on some internal, departmental websites where branding and skinning will never in a million years get funding. On the other hand, customer facing sites, such as car manufacturers and department stores live in a world where every single new product that gets launched results in an entirely new skin for the website. New colors, new layout, new background images and all of this lead by designers who should be able to make the change purely in css so there's no chance they can break any working php (or what-have-you). There are also branded sites, where you have multiple skins, potentially running on the same site simultaneously. On sites with that requirement, you can't have visual changes impact html or you end up breaking every other brand just to update one of them. In these situations, semantic class names are a necessity.
The second thing that often gets left out is how to combat the problem of repeating groups of properties created by semantic class names, as in:
.content-nav {
float: left;
margin-right: 10px;
background-color: #ccc;
}
.content-nav .user-photo {
float: left;
margin-right: 10px;
border: solid 1px #000;
}
.content-nav .user-display-name {
float: left;
margin-right: 10px;
text-decoration: underline;
}
People often point this out as a drawback of semantic names, and I think that that's a valid point. On the other hand I would like to point out that there are tools that can help you keep semantic css DRY, such as LESS and SASS. I did see one other commenter mention this above, but I just thought that this point was worth highlighting.
I think very often tabs are implemented as <ul> and a series of <li> inside. Is there advantage of that over
just using <div> with a few <div>s inside?
Usually with <ul> as tabs, the padding-left of it needs to be reset to 0, and list-style needs to be set to none, while <div> doesn't have this issue.
Short and clear answer: Why should I use 'li' instead of 'div'?
Excerpt:
"For semantic correctness. HTML has the ability to express lists of things, and it helps the Google robot, screen readers, and all manner of users who don't care solely about the presentation of the site understand your content better."
"For the visually impaired, it can be helpful to distinguish what's in a list and what's not. Say if you have a list of ingredients in a recipe for example, and the user wants to skip to the instructions or just read the list, you need a list."
Don't forget to visit the link to learn more.
Looks better when CSS isn’t applied, and is (I believe) easier to interact with in screen-readers.
In terms of adding/changing the code, it is more readable and makes more sense. Tabs/navigation naturally seem like a list where the items are associated with each other.
Once implemented via CSS, it is extremely easy to add/modify when the time comes to change your navigation/tabs.
How would I go about testing the performance benchmarks of different css selectors? I've read articles like this. But I don't know if it is applicable to my website because he used a test page with 20000 classes and 60000 DOM elements.
Should I even care,does performance really get downgraded that much based upon the css strategy you take?
Fo example, I like doing this ...
.navbar { background:gray; }
.navbar > li { display:inline;background:#ffffff; }
<ul class="navbar">
<li>Menu 1</li>
<li>Menu 2</li>
<li>Menu 3</li>
</ul>
... but I could do this ...
.navbar { background:gray; }
.navbar-item { display:inline;background:#ffffff; }
<ul class="navbar">
<li class="navbar-item">Menu 1</li>
<li class="navbar-item">Menu 2</li>
<li class="navbar-item">Menu 3</li>
</ul>
Some would say (and might be right) that the second option would be faster.
But if you multiply the second method across all pages I see the following disadvantages:
The page size will increase because all the elements having classes
Number of css classes can get quite large which would require more css class parsing
My pages seem to be ~ 8KB with ~1000 DOM elements.
So my real question is how do I create a test bed where I could test performance deltas based on strategy taken for realistic web page sizes? Specifically how do i know how long it takes for a page to be displayed? javascript? how exactly?
Help and just plain opinions are welcome!
Check out the Page Speed extension for Firefox. Once you run it for a page, under "Use efficient CSS selectors" it gives you a list of the offending CSS selectors along with brief explanations.
Also, there's another extension for Chrome - Speed Tracer. Amongst other things, it offers insight into time spent on CSS style recalculation and selector matching. This may just be what you are looking for.
From reading the article you listed it looks like the difference between the two type of selectors is not worth worrying about. Make certain the css is clear enough to understand it, and only worry about speed after that proves to be the bottleneck.
there really is no need to do
Menu 1
you can have a css class
navbar li
remember too that external css files can be minified and cached whereas the html cannot. Performance is also a relative term: do they visit frequently? Is the network slow? are they state employees with ancient computers running IE6?
I don't have a direct answer to your question of how to build a page speed testing program. However, I will offer you some best practice guidelines to follow that will steer you in the right direction.
The page size will increase because
all the elements having classes
The size of adding numerous classes is very much negligible for two reasons:
1) The additional size in your stylesheet will be cached (assuming you use externals, which you should).
2) The HTML markup from adding numerous classes to the DOM is 1kb at most. If it's more, than you need to take better advantage of inheritance.
Number of css classes can get quite
large which would require more css
class parsing
You will have more CSS classes, yes... but binding rules to a CSS class is actually faster than the alternative of using descendant or adjacent selectors.
As long as you avoid descendant/adjacent selectors as much as reasonably possible, you use external stylesheets, and you take advantage of inheritance to reduce duplicate rules, you shouldn't need to worry about CSS performance stress tests.
You're creating an HTML layout. Let's assume that you don't need the benefits of multiple stylesheets, that a small increase in HTML size is not a concern, and that you have a style which will only be used once. I'm often in favour of using an inline style here, as I view the repetition of your CSS class name or ID as the cost of an abstraction you don't currently need, and may not ever use.
Standard doctrine these days is to always create HTML layouts using semantic markup and CSS styles, so am I missing something here? Please let me know your thoughts.
Even if you only use a particular style once there are still benefits to keeping it with your other styles and not putting it inline. First, there is the separation of concerns that leads to improved maintainability. If you know you are going in to make only a style change, there is a single place to look for any changes. Another benefit is the self-documentation from having to type out the class name. By giving that style a name, even though it is used once, it makes the semantic code below more declarative -- you can read that not only is this random p a paragraph, it is also, say, the intro paragraph.
This is, of course, assuming that you are never going to use that particular style again. If you might than there is even more reason to factor it out into a named style. Inline styles aren't evil, but they are somewhat of a gateway drug.
Ideally your CSS should be "Object Oriented" (at least, as OO as CSS can be). You should "inherit" from classes that set common properties and create new classes when you define properties that could be used elsewhere.
Take a look at the OOCSS project which is trying to espouse these principles (or re-introduce them as it were).
To quote Welbog:
... It seems to me that "OOCSS" is just CSS that isn't written haphazardly. Much the same way you can write non-object-oriented designs in OO languages, you can easily mess up the fundamental ideals upon which CSS was created. OOCSS seems to be saying, "Let's not screw up anymore, guys."
One advantage of keeping the HTML and CSS separate is that you can re-skin the webpage without changing any of the HTML.
Steve
There are some situations in which I usually neglect creating a new class for a simple style change on a single element. It is usually pretty clear when you are doing it that there's a low-to-zero chance of you needing to apply that particular style to something else later down the road. The most common case for me is when I need something to have a particular padding/margin to be in the right place but it's not an element important enough to have its own ID.
This may not be a popular opinion here, but in those scenarios I don't think an inline style is evil.
Personally, I've found that I have an element or two and I would put an inline style in, go back and see that I need more than that element, so I'd change it to a class or forget about it and be not able to change it.
You could also try putting a certain div / page class, and write descendent styles for that in the stylesheet instead of inline elements.
Also, if you ever decide to add javascript, you won't already have a well-labeled class there and you'll need to change it.
Usually this isn't much problem with dynamically generated websites, but it can become a large problem when you go overboard and have tons of inline tags to switch out. It also can make it harder for people if they wish to disable styles for accessability etc-- you usually can overcome this by using a class.
Say, using <b style="color:red">bold</b> instead of body.products div b {color:red}.
I'm personally a fan of selectors, not using too many classes. They are more reusable, and you can edit the whole site in one place with the stylesheets.
But this is overkill <p style="font-weight:bold;font-size:1.2em; text-index:20px;">Indented Bold Paragraph</p> so it this <p class="indent bold larger">text</p> instead you can door ``<p><b></b></p>.
"A foolish consistency is the hobgoblin of little minds"
So, in this case is which is the foolish consistency? :) Why does DRY take precedence over the separation of markup and style?
Can you be sure that your CSS rule will only be used once? More over, how can you be sure that it won't need to be changed in the future, and how can you be sure that you would be the person needing to make the change? Are you sure you even need to add a class or id to target this unique element?
I guess I am having trouble seeing how adding
<input type="submit" style="border: 1px solid red;"/>
is some how "superior" to 12 or so more characters
<input type="submit" class="b-red">
.b-red {border: 1px solid red;}
or to a potentially equivalent character count
input {border:1px solid red;}
Of course there are situations where every rule of thumb can and should be violated. The question is, what do you gain from following DRY that outweighs the importance of following markup/style dichotomy?