I was thinking about what weenog suggested above...
For BAB to really shine (though still nowhere near as much as full casting), it would need to be the first and final word in combat maneuvers, damage output, and number of attacks. A barbarian losing a grapple with a level-appropriate undead (2/4 BAB) should not even be a possibility to consider, even if it is 3 sizes larger and has a Str score in the 40s. A wizard's best option for solving a problem that needs to die should be hiring on a knight or the like and pointing him at the problem (none of this heavily metamagic scorching ray, or save or die spell type of stuff). The fighter who TWFs should be making at least half the attacks per round (counting both sides) in any given combat that doesn't include another 4/4 BAB class, with most of them at or near his full attack bonus.
I sat and thought about it for a bit. I don't think creating and enviorn where its impossible to defeat melee's in melee is what is desired either. I hate to contrdict myself, if I ever stated the contrary, but I don't think creating a scenario where bab is greater than casting or whatever sucks is any good either. I'd like a system where both of them are parsed to shine when you choose them.
Perhaps a system where the number of actions was governed by BAB? Or being able to burn an iterative to take another standard action? I'm not quite sure how to balance that out, but it seems to me like the only thing close in power to spellcasting is action economy. Give melee types an advantage in options they can take.
Well, the key factors here(as discussed just before the damned derail), was that BAB does not do its job at making you better at fighting. Combat ability is more than just accuracy, for BAB to count as a basic statistically significant thing, lets take a level 10 warrior character and leave him naked. Now lets take another warrior, at level 5 and put them in a cage match. Give them both a regular club, caveman style.
To hit - The level 10 dude has +15 to hit between str 20 and BAB 10, the level 5 dude has +9 to hit between str 18 and BAB 5.
Damage - Both of them being naked, the level 10 warrior hits for....1.5 more damage than the level 5 dude's 1d6+6(9). Good thing he has an iterative attack, so he hits for twice the level 5 dude, 2d6+12(18) in all.
Health - The level 10 dude has twice as much health. Going with Con 14, the level 5 version has 5d8+10(36) and the level 10 has 10d8+20(68).
AC - They both have the same AC of 12.
So discounting even the effects of spells, combat style and gear, just a straight out beef to beef contest, BAB and HD doesn't do a whole lot does it? You could replace one of them with a rogue of the same base stats and not notice much of a difference. Or a sorc who's out of spells even(with a mysteriously 18/20 Str). The fight would go similarly, even though its intentionally set up to maximize the visible effects of a higher BAB(the extra attack vs no extra attack).
Thoughts on the effects of making BAB grant to hit(+1/BAB), AC(+1/BAB) and damage(+1/BAB) bonuses? This makes it more significant than any single stat, as well as most buffs. It does make rocket tag worse without a corresponding DR improvement, but we'll leave that point for later.