This sounds like an ambitious project. I'll be watching this with interest, but not joining in, I think. I'm attempting much the same, but I tend to go for a little more complexity than the base rules, rather than less.
Just as a discussion point, on skills, what I do in my CityScape game (which is d20 derived, but far-removed) is I have a base Combat bonus, and a base Special bonus. Skill ranks (which include weapons skills) grant abilities or act as their prereqs, but the skill rank total is compared against the appropriate base bonus, weapon skills and such against the combat bonus, utility skills (and spells) against the special bonus, with the lesser of the two totals being the actual effective dice bonus. (this also allows me to stack two requirement abilities for ability-effectiveness, running them against each other so a large bonus in one doesn't overshadow the other) Much more complex, but also easier to control effectiveness and determine mechanical flavor in an otherwise loose-structured system.
I've been thinking that my own D&D-fantasy revision, which is more like 3.5e & 4e, which I've been calling 3.9e, might be borrowing in more and more from the CityScape concepts, just because they seem more functional at times, albiet wrong in flavor. (in CityScape, magic isn't divided into arcane and divine, and everyone uses magic, just different non-exclusive subtypes, whereas my D&D would have Arcane, Divine, Primal, Eldritch, and a little bit of psionic.)
What I'm thinking I might do for my 3.9e is simplify the BAB into full and half; combat-centric characters get full, magic-centric characters get half. Then, individual caster-level is likewise skewed to half or full, but on an unlocking basis, so a mage20 would have a +10 BAB and a full arcane caster level, while a knight20 has a full BAB, but no arcane caster level. If that knight had 1 level of mage, he would instead have an arcane caster level of 10. (As an aside, every class in my game will actually have a half caster-level of some kind, because it's flavored as intrinsic magical ability everyone attains with levels. Even rogues will get psionic caster level as a backtype, and psionics is otherwise nonexistent in my gameworld.)
Why I mention all this is, I might borrow back my own Special bonus to use as a skill-effectiveness limiter, to differentiate int-heavy casters from int-heavy skill users, and I've been debating putting weapon skills and magic into the game, likewise using class-derived caps to establish field dominance. Types can be played against, and feats would allow the system limits to be broken to favor a concept (like a gish) but the system would be more directed. For skills, a class-skill-trained bonus like PF's (but also growing with levels) is added to the ranks, and compared against a Base Skill Bonus, and the lesser of the two is what's used on the d20 roll. Classes with less than full BSB have less need for a high number of ranks per level, and could spread themselves out to different types of skills each level gained. This would allow a class to have a wider class skill list, without requiring them to take every level at every level opportunity.
I'm also not joining in because I like using non-SRD mechanics & concepts too much, so my D&D homebrew will have to remain for my own personal use, rather than as part of a sellable product. (my magic system in CityScape is much more original, but also tailored to my Nightmares in the Dark setting.)
(Edit: And the purpose of this long-winded meandering post isn't to try to steal the conversation, which I just realized it looks like, (sorry), but to throw out my own system concept for your examination, and if it looks interesting, offer it up for your use. I like it, but I'm biased.)