Ironically in the context of the original article, it's precisely when you're trying to support IE 5+ or 6+ (or even more modest goals) that less/sass is really valuable, since you can generate chunks of CSS values with all the browser prefixes/hacky workarounds/etc, and still be able to think about things at a higher level.
I think "some code duplication" massively understates it. Being able to use named variables for colors is on its own enough of a win to make up for whatever truth there is in the "harder to train developers" downside. The many other wins make it a no-brainer.
Re:The boring front-end developer
by robinson7d on Monday July 13, 2015 @02:15PM
(#9879892)
Agreed. One example would be that using named variables, particularly set with "default" values, makes white-labelling applications much simpler.
It's not the only downside to consider. http://adamsilver.io/articles/the-disadvantages-of-css-prepr...
There are a lot of advantages to them that need consideration too granted.
Certainly, everything has trade-offs, but CSS pre-processors are more heavily weighted toward the positive side of the trade-off calculation than pretty much anything else I can think of, especially in the "web technologies" space. I found lots of the arguments in the OP to ring true – things often feel far too fancy, and the more "boring" stuff has major appeal – but I found the arguments against CSS pre-processors there and in the article you linked very thin relative to my experience with them after years of using raw CSS.
I'm sort of coming around to this perspective as well. Ever since I read this article[0] on more of a APL/J/K style to CSS (the author's blog's CSS looks like this[2]), I've noodled with the idea of doing straight CSS again, in this pretty condensed format. It's kind of shocking at first, but does have its own charm.
[0] http://prog21.dadgum.com/200.html [1] http://prog21.dadgum.com/p21.css
Just look at the LESS example on the front page. The compiled output is simpler than the source you're writing. I find having some code duplication to be an acceptable trade-off for turning a basic declarative configuration language into a chimera DSL that must be incorporated into the build process.
The more a configuration language strays away from being declarative, the less desirable this becomes.