The new Legends & Lore post from Mearls talks about the concept of Class Groups, as well as some of outbound design ideas behind the Mage class in the last two playtest packets. Kainenchen has already written a post about this topic, so by all means go read her post first. For my part, I think that ideas are getting weirdly conflated here, and if WotC wants to think in categorical terms, they need to set the boundaries of each category in a more consistent way. Beyond that, I want to talk about the concept of classes in the first place.
For those of you who haven't clicked either link above, the deal here is that they're looking for a new way to, er, classify their classes - to figure out which containers to put each class in, partly as a design guideline. In 4e, they handled this with Defender, Striker, Leader, and Controller, a system dynamic that was pretty acceptable to people who liked 4e but a major point of contention among those who didn't. Personally, I never understood the objection; these aren't classifications that force anything on the player. They just tell you in advance what you're getting when you sign up for a particular class. With particular builds and customization choices, it's possible to nudge a class toward something other than its native combat function, and classes generally have an obvious secondary role anyway. The argument, though, that you want to play a fighter who is more about dealing damage than about protecting allies is a perplexing one in the 4e design space, because it doesn't know what to do when you care more about the class name and your own mental associations with it than you do about the presented function.
For D&D Next, they're instead reverting to the classifications of 2e, changing only the Thief/Rogue category to Trickster. Their class design, however, has little to do with what it was in 2e: in those days new class abilities were rare and customization within a class was almost completely unknown, to say nothing of the total absence of any concept of feats. The 2e model wasn't called upon to handle a lot of the class archetypes that D&D Next will be: swordmages, avengers (I hope they support avengers), warlords (probably an option internal to the fighter class, which I want to say more about in a minute), and so on.
Okay, so you can say that swordmages and warlords are warriors, while avengers are either warriors or tricksters, but we're getting into seriously eroding the utility of the classifications. The swordmage (any combination of melee combat and arcane spellcasting; a particular focus on defense is nice, but optional) is a concept that has been late to the game in every edition of D&D I've ever played, but it has appeared - the regrettable Bladesinger of 2e fame (or just the elven fighter/mage multiclass), the duskblade of 3e, the swordmage of 4e. It doesn't really make sense to say that the priest is where healing comes from, when bards, paladins, rangers, and mages with the good sense to splash one level of cleric or druid do almost as well. Other than healing, does anything set divine magic apart from other magic?
To put all of this another way, this class-group structure is halfway about function in combat and halfway about power source, and getting the two mixed up looks weird. What I suspect they really want to do is to define Might and Wits as separate power sources, just as Arcane and Divine are separate power sources, and think of each class as having at least a primary and secondary power source (possibly more). This would actually help one of my quibbles with 4e: I didn't like that all of a paladin's attacks were nominally magical, and the paladin didn't do anything (aside from basic attacks, which he has no reason to make) that emphasized the character's physical, fighter-like prowess. Similarly, if my recommendations about making Wits its own power source were incorporated, I'd hope that fighters and rogues both had good access to Might and Wits to support the thuggish kind of rogue and the swashbuckling kind of fighter.
So about the warlord thing. A warlord is certainly a type of warrior, in concept, and leadership features are a great way to expand the fighter class's concept to be something more than a brute with a sword. That's the good news. The bad part - well, inconsistent - is that it makes the fighter a very broad class concept, while other classes are very narrow. There's no particular reason that any two fighters would have all that much in common, while even the most different of paladins (going back to a packet that had variation) are as similar as applying different color palettes to the plate-and-maille-covered figure.
This brings me to the post's discussion of the Mage. Warning: This is going to get technical. See, a couple of packets ago, they changed the Wizard to the Mage, so that Mage could be a container class for every kind of arcane spellcaster (except for, apparently, the bard): the sorcerer, warlock, and whatever else could all become decision points within the Mage class. They discovered that this approach is unpopular, and while I can't be certain why other people don't like it, I think it's worth discussing some of the more apparent strengths and weaknesses of that container-class approach.
The strong points, which WotC points out, are that if X is usable by a mage, it's usable by everything in that container: the spell list, itemization (though I think there are better ways to carry class-restriction criteria), feats (though D&D Next really doesn't need class-restricted feats in its current model), and so on. I've probably had two dozen separate conversations with Kainenchen and Stands-in-Fire about the strengths and weaknesses of container-class design, or few-classes design as we were thinking of it. The other strong points have to do with conceptual clarity and the greater engagement of making fewer, more significant choices as opposed to a larger number of less significant choices.
On the other hand, many of those strengths carry weaknesses as well. If wizard and warlock are both part of the Mage container, the mechanics that let you choose something from two containers (multi-classing) don't apply cleanly, because that would make you a mage/mage. If that's what you prefer - because for some cosmological reason in your setting a mage can't pursue the Arts Arcane and the Left-Handed Path - then you're set, but that kind of cosmological assertion won't work for all campaigns, and these rules are supposed to be as universal as possible. A simple statement as part of the setting that X and Y separate classes are mutually exclusive would be a better way to store the information that one person cannot do both.
If the use qualification for an item or feat is the container-class, you still have to go back and review the applicability of those things to any new Path or subclass you want to create, or have a new way to disqualify a Path from a particular piece of content. To use a 3.x example, Rary's mnemonic enhancer was wizard-only, where basically all other spells that the wizard could use were also available to the sorcerer. Once you have to worry about special exclusions, blanket inclusion isn't nearly as much of a benefit anymore. Further, once a Path is mechanically distant from other Paths within its container class, it becomes just a messy way to present information.
Now, they could go back to a much simpler design approach, supporting "paladin" as an outcome rather than a class unto itself, where the player would presumably mix fighter and cleric to some percentage or another to represent a warrior with some access to divine magic. That's a fine way to do things, but it is absolutely not what they're doing at this point. They want to create mechanical interest with new classes, splitting off still more specialized concepts of those classes. You know, Mage-Wizard-Evoker, Ranger-Horde Breaker, and so on. That's also a fine way to do things, as the massive success of 3.x and derivative products suggest, but D&D Next has steered toward simpler, more direct mechanics to get the job done - but that's more like few-classes than container-classes.
The article ends with a suggestion of what they expect the class-groups model to accomplish for them: classes in the warrior group have d10 or d12 HD, tricksters and priests have d8 HD, and mages have d6 HD. I am in favor of narrowing the range of hit points, especially in a game without explicit tanking mechanics so that the warrior-types can protect the mages. It looks like they're still nailing down what the monk is supposed to do, trending toward making them warriors rather than tricksters - though if previous models were any clue, that means they'll be warriors with better skill access and basically no use for treasure.
I can only imagine that endless criticism will be heaped upon them no matter what titles they choose for class groups. I'd like to suggest that what they really want are warrior, trickster (or expert), pure-caster (may or may not involve healing, but definitely involves spells or comparable ideas like psionics as the primary tool), and mixed-use caster (some weapon-based combat and/or skill use mixed in with casting - bards, clerics, avengers, and so on). The Hit Die model they're pursuing still applies here, without the context-laden terminology of "priest" and "mage." It leaves the door open to a WoW-like cloth-wearing, spell-casting healer-mage as a class design (rather than a player consciously playing sub-optimally), something that D&D has offered only rarely and reluctantly before (here going back to the archivist in 3e's Heroes of Horror). I'm not sure why it's this hard for them to just break things down this way, then derive new classes from that class-type plus a power source; NERO has operated off of a version of this model since 1988.
Actually, I do know; they're beholden to their classic core-four classes. I am fine with D&D sticking to the paradigm that made it the most played tabletop game in history; I just think that the core-four should be placed inside the class groups once they're named and designed, rather than the class groups being shaped by the classes they're trying to describe. The container has more room if it isn't molded around the first thing it contains.