...I am a bit angry about cross-classing. Sure, 3.5 had far too much, but making up for it with far too little is no answer. Multi-classing should not be an essential aspect of character design, but nor should it make the character sub-par. I can't claim I know exactly how to accomplish this, but if multiclassing is done through feats, make each feat for the cross-classing abilities just as good as any other feat someone might want to take. Maybe allow a new class feature to be purchased via feat from the class you're multied into. That's what my house rules do.
That's kind of how feats and skills developed. In 2nd edition, there were "Weapon Proficiencies" and "Non-Weapon Proficiencies". The latter became skills, the prior became feats, with third edition. I can't claim I like it too much when the mechanics of the game get involved with role-playing. I consider it a good idea to get rid of most crafting skills, for example. It makes no sense for a 20th-level rogue to be a supreme chef through fighting monsters, and the level 10 commoner chef (a damned high level, considering the dude's a commoner (3rd edition rules, btw)) , whom spends his entire life as an apprentice chef and then a full chef, is a less good chef.
There are more than a few days when I get this unshakeable hunch that D&D would be well served to dust off some of the Dragon Quest class division ideas. I say this in full awareness of some of the drawbacks to going that way.
In DQ, classes are called "professions", but then magic (spells in schools) and weapon ability is bought separately, through something closer to skills. However, unlike a skill-based or point-based game, almost everything you buy is treated like a class track. Some are broad and some are narrow. When you buy broadsword skill ranks, you do in a narrow track that applies to ... broadsword skill use. When you buy ranger skill ranks, you get several thematically related abilities in that profession. XP is spent on these things, and they get progressively more expensive the better they get.
So I can envision a version of D&D where you spend XP in such tracks, separately. If you want to be a 10th level ranger with 2nd level broadsword ranks, you can.
The obvious objection to that is one that can be easily answered. We are so used to the cherry picking in skill-based or point-based games, that the DQ solution gets short thrift: Make XP award partially contingent upon broadening your character. I forget the exact details, but DQ RAW had three sizably different XP tiers, and you had to get 8-12 professions, skills, or schools to set ranks to qualify for the tier. Once you made it, you got enough XP to make creditable gains in those tiers. If you over-specialized, you can get rank 10 broadsword use very early--but you are seriously hurt elsewhere (and it matters, even for fighting). OTOH, If you spread out too much, it takes forever to get that last rank in the 8-12 skills.
In DQ, it was hard-coded. But DQ is a very old game. It would be interesting in D&D to have such tracks for advancement, but then have multiple recommendations for how to structure the tier thresholds to get the kind of game you want. For example, if you want something akin to 4E, where every character has to work hard to not be about equally good in combat, you simply set up the tier requirements to require a certain amount of combat ability. If you want a lot of slowly developing generalists, akin to say, GURPS, you relax the requirements severely. If you want some specialization, but not too much overt power-gaming, you go for something more middle of the road.
Another nice thing about this structure is that you need not fill out every track equally. You can lop off tracks for thematic, simulation, or gameplay reasons. If you don't want 20th level cake bakers, then you don't let the track with baking skills go very far.
Edit: Yep, WotC does have the DQ license. TSR got it from SPI, put out DQ 3rd ed., and then sat on it. WotC got it when they bought out TSR.