Just let me throw this in, when talking about what 5e might become. On our blog someone wrote that he things 5e might get a modular approach. One very simple core system (like Mearls musings are) and optional packages for further additions, like skills and whatever. Or even, and that sounds much more tempting to me, doing all the specific stuff for different campaign settings as modules. Like a set for modern, another for Eberron.
I find the idea of a common subsystem and additions for settings pretty interesting, as long as e.g. the fantasy settings will not differ totally. But the idea of having optional rules for everything is pretty scary. Sounds like every single group is playing by different rules then and you have to negotiate rules before starting even to build a character.
Modular does not necessarily mean more complex. In fact, making something modular is a good way to reduce overall complexity and specific complexity (i.e. the complexity that you actually deal with at any given moment).
I'll grant you that in the context of, "filling page count," modular in the past has often meant, "long list of options, some of which don't work very well or don't belong on the list, and thus create all kinds of trouble." So history gives reason to be cautious about the attempt.
The trick is providing modular hooks at key, important points--and not all over the place simply to be modular. Object-oriented software developers learned this 20 years ago (after over a decade of fighting with it): You don't make everything so that you can swap it out. Many things, you want them to be one way and stay that way. All that fussing around trying to make everything modular just make it difficult to swap when you do want to swap. Instead, you pick those key things that are supposed to be modular--and then you make them incredibly easy to understand and swap.
An easy example in D&D might be the complexity of the weapon lists. You might have a base option Weapon B in the base game, much like the Red Box list, where you have 20 or so of the most iconic weapons, and damage ranges and abilities don't vary much. Then you have ultra simple option Weapon S where weapons are just color, all do the same damage, and about the only changes are in melee versus ranged. Then finally you have complex option Weapon C where the list is rather comprehensive, with real difference in mechanical effect, divided into groups, etc.
If the designers decide this is important--
if--then the rest of the system touching weapons needs to work with all three versions. And you don't blend the versions (short of a player perhaps using Weapon C to get an idea for his color while using Weapon S). You don't, for example, have certain weapons in C doing greater damage to "large" creatures, since that effect is not easily replicable in S and B without skewing expected results. (That is, you don't design your creatures such that you need this effect in weapons to make the whole system work. You might have something analogous in flavor in the weapons which doesn't touch the monsters themselves--e.g. how many hit points they get.)
Like, "exception-based design," it is not going to work to say, "modular-based design," and then have everyone go wild with exceptions or modules. The art is in the choosing.
