Just to be clear, I like having some niche classes. Where I have my doubts is the idea that the archetypical ranger, paladin, bard, druid, barbarian, assassin, etc. are the best use of those slots. That is, I'm not against any of them, per se. But I am against any of them that are kept without due consideration for the things around it. The reasons, good or bad, are thus different with each class, and with each edition.
For example, a good reason to have a paladin class is that you've got this warrior type over here and this priestly type over there, and you want to stake out some other ground that is somewhat of a mixture but also brings its own elements into a nifty syngery that is greater than the parts, and "class" is the element of the system best positioned to take that ground.
A bad reason to have a paladin class is because we've always had one, and no matter what happens with the fighter and cleric during the course of the design, we are gonna squeeze a paladin in there somewhere, if we have to get an industrial-strength press to make it thin enough to fit--or radically change the system to leave it a spot.
In reality, the decision will never be that pure, either way, of course. And I don't mind a great deal of soul searching, bending, and even carefully selected system mangling to make the big, traditional ones work. (I'll extend a lot more room here to paladin, ranger, bard, and druid than I would to other classes, too, because they have a bigger claim on that tradition.)
Conversely, if having decided to make most or all traditional classes fit as a goal, then it becomes incumbent on the designers to make a system that readily accommodates them. That's fine to, since this isn't so much a "tail wagging the dog" thing in D&D as a, "who steps first, left leg or right?" thing. That takes us outside the realm of "good reasons" to have a class, though.