Example: <div class="header inset aboutUs"> or <div id="header" class="inset aboutUs">
The difference is the second option shows at first glance what the element is supposed to be representing to someone reading your code.
> if you're only going to be using the style once on a page
Herein lies the the problem. You now can't use that header anywhere else, meaning the it's non-composable. You end up writing less, and more maintainable CSS when you treat components as immutable and composable.
Most people consider the additional readability of separating out the `id` to be negligible.
You can use that header on any page and you're not going to re-use it on the same page. Why does it need to be "compose-able" at all?
Either way this is a silly argument because there is no problem here. You have a set of CSS styles you want applied to an element on the page that will never have a copy, etc? You can use ID or Class and you can even swap them with the same level of difficulty. This is entirely a non-issue no matter what way you go; it's more of a style thing.
#login-widget {
@extends widget();
}
.login-widget {
@extends widget();
}
If that somehow sits fine with you perhaps consider the rule the rest of the software world has accepted long ago -- 'this code will be changed'. The fact you think you will really only have one of these things on a page at once (hence making it suitable for an ID) ... that WILL change.
There will come a day when you have a login-widget in your sidebar, yet at some obscene last-minute moment the customer will demand a dedicated login page too, your developers will want to use your base template (which has your login widget visible in the sidebar) creating 2 login widgets on-page, cursing your assumption that you thought you knew better.
Ignoring that, you could forget my example above (I'm sure there's a way you could dismiss it as contrived) and learn a lesson from the OOP crowd and just never write singletons [0] ever. It's just a really bad assumption to think you'll only want one of a thing.
I couldn't disagree more. The more specific you are the less side affects you'll encounter in the rest of your project. When a project gets incredibly large if your styles are NOT specific you're going to run into a lot of unintended consequences.
This doesn't mean I encourage using IDs all over the place; they should only be used when there can be a single version of that element on the page. But being specific also means using the cascade (e.g. html > body > div.mydiv versus just .mydiv).
You can handle 99% of use cases for specificity just nesting classes. Using IDs to settle specificity is the same as bringing an ICBM to a Nerf gun fight. When I was doing front-end development, I spent more time fighting back against overly specific CSS, thanks to IDs (and fellow developers who just sprinkled `important!` on everything until it worked), than struggling because I couldn't get specific enough without IDs.
Nesting classes can get really messy and I'm not sure that really makes sense to do in "99% of use cases" if the goal is to make something specific. Mixing classes allows you to apply multiple styles onto a single element which can be really handy but using that as the way to target a specific, single element means you can't re-use some of those classes in the same way otherwise you run the risk of targeting more than one element in the same way.
Specificity should usually be handled through the cascade and / or IDs; that's what they were meant to do. Classes provide buckets of styles as do IDs and both can be used in the same way with IDs targeting only a single element and not a compose-able one.
Ultimately you can specifically target in CSS a variety of ways. Outright saying one way is just wrong or "[an] ICBM" kinda ignores the use cases of the various ways to do it to attempt to force a single answer where I do not think a single answer is satisfactory.
I try to reuse some style to only find it doesn't work when placed somewhere else in the HTML.
Having too generic styles is also a problem though. E.g. styling a label element on the assumption it's in a form can be frustrating.
It's all about the right balance and styling specifcally what you mean, not the specific structure of the HTML.
Not sure why you'd make such an absolute claim. When you say single version are you saying single html element or single CSS definition? If the former then I disagree (headers, footers, main containers of content, etc), if the latter then I don't understand how this would be the case. In HTML IDs are limited to one specific element but you can refer to it as many ways as you want in CSS which means you can use different media questions, handle resolutions, print, etc.
So I'm not really sure I understand your point.
> Being less specific allows your css to be more manageable to change in many different situations. In complicated layouts, cascades are very non-friendly.
I've worked in a lot of really large projects and I'm not sure how little to no specificity could be more manageable due to side-effects. Ultimately you're going to need a mixture of specific and non-specific styling so arguing whether you should be more or less doesn't really make sense to me; just use the right level of specificity for the job at hand. If you have a page with specialized content you might want to be more specific there to not mess up another page. If you have a lot of re-usable components you probably want less.
And you might run into a circumstance where you have to repeat the header or footer -- perhaps you're allowing someone to edit the header or footer of your site and you want to do so in the middle of the page? Or you want to perform a complex animation that would require having multiple elements on the page at the same time. IDs ultimately are more brittle, and they don't do anything to enforce that they're only used once on a page. Their best use appears to be for hyperlinks or perhaps actual ID numbers (e.g. JavaScript) and not much else these days, if you're being consistent.