This is my take on the 3.0/3.5 split.
3.5 is a tighter rule-set, but it began a drift toward "innovation" over "tradition". 3.0 was VERY firmly in "new system, nostalgia included" syndrome from Greyhawk in Core to the Great Wheel in MotP, etc.
3.0 had 16 base classes (11 PH, 2 psionic, 3 oriental) but dozen's of PrCs. 3.5 seemed to reverse this trend
3.0 wasn't entirely balanced and some combos were OBVIOUSLY better than others, just like 1e/2e, etc. When things were good, they were GOOD (Haste, Harm, Gspell Focus) and when they sucked, they sucked (Bard, Ranger.) 3.0 had not taken into consideration the extent of changes the new multi-classing system and feats would have. Thus, there were some real wonky elements and some downright broken.
3.0 never strayed too far from the core. Almost all 3.0 books only ever referenced the Core Rules (and the FRCS, if a Realms book) but rarely did they mention anything outside the core without reprinting it (hello long list of feats in Deities & Demigods).
3.0, after its initial release, didn't rock the boat with many new rules (beyond the psionic system Epic, and Divine Rank). No warlock invocations, no knights challenge, no powerful build, no Swift/Immediate Action sidebars, etc. The three mentioned 3.0 additions (psionics, epic, divine rank) were also very easy to ignore, whereas swift/immediate actions are not (if you use things beyond the core.)
This created a feeling of a system which seemed more contained, but allowed power-levels to fluxuate wildly. It felt truer to the D&D of yore, but failed in its attempt to "level the playing field" like it had wanted to. 3.5 did a better overall job of balancing the core (overbalancing it in places) and moving the game in a new direction (for good or ill).
