Learning Linux is like learning LaTeX. It requires serious effort to become powerfully useful but then you never want to miss it because there is nothing that can actually compete.
OSX and Windows are easier to handle at the first place but if you get into the details then you may have more trouble than in Linux (driver problems, document formats) because there is no obsolescence in Linux. In OSX and Windows you depend on the cooperation of the manufacturers. In Linux drivers, applications, and document formats written decades ago still work.
> Learning Linux is like learning LaTeX
This is a particularly fantastic analogy when you look at the type of error messages that LaTeX produces.
Nevertheless LaTeX is still the favorite DTP system among scientists. LaTeX is actually not too hard to learn. It just requires some mental effort to get it. Scientists are willing to do that. Some people obviously don't.
Maybe but using a Linux distro like Ubuntu or Mint or similar is more like using an Office package.
I used to make this claim as someone who had been using desktop gnu/linux for several years. I've been using OSX now for a year and so stopped making the claim. But it is really frustrating to hear people say "it works for me" to rebut claims that it doesn't work for most people. Such counters and arguments over them are basically single data points. If the point under discussion is "does desktop linux work well enough to gain massive amounts of market share?", then you need to run actual large scale trials comparing how hard it is for people to...
1) Start up a spreadsheet program and do things
2) Check social media and play some random video
3) Play League of Legends
4) <insert your favorite thing here>
Some of these things are indeed comparably easy in, say Ubuntu (or were pre-unity) for me, but I was an MIT undergrad that hung around SIPB, the computer club. My massive amounts of tech privilege make it impossible to individually evaluate if the user experience someone has with spreadsheets is acceptable unless I put real study into it. The same is probably true of many HN commenters. So until someone spends the time and money to run actual user experience studies, we have no data to say that the linux desktop can take over the world.
Also, there isn't even one "Linux desktop". There is mint, Ubuntu+gnome classic, Ubuntu+unity, ubuntu+some weird xmonad setup.
Another argument I often see for why Desktop Linux has an advantage is that "Linux is free as in beer". But it is not. If someone making $15/hr takes more than 8 hours (not unrealistic, at least for an MIT CS student back in 2010) fiddling with things and trying to get them to work, then desktop Linux is more expensive.