Diplomacy and other charisma-based skills in the 3.5 Dungeons and Dragons rules mark a real departure from earlier versions of the game, because they encourage storytelling in a new way -- well, new to D&D players, anyway. Under first and second edition, non-combat interaction was handled arbitrarily, through pure role play and perhaps a few impromptu charisma checks. This was fine for extrovert players with real-world acting talent, but it left the shy and socially clumsy players out of the loop. Chris Magoun wrote a good essay on this topic.
Sure, the role-play purists out there may scoff at reducing character interaction to a die roll, but in my opinion, if using a dice-based mechanic gets people role-playing who otherwise would ignore that aspect of the game, then great!
But even so, all is not peaches and cream. The d20 mechanics for resolving diplomacy, bluff, intimidate and other social conflicts are problematic. It is too easy for a dedicated d20 diplomat to pile on the bonuses -- and that is problematic for a system in which the difficulty checks do not scale. The problem gets worse at high levels, as described in this spirited DnD-L thread on Bluff. Just about everyone thinks the system is flawed.
Just ask Mike Mearls, who recently blogged that Diplomacy is B0rken. Mearls says, convincingly, that it is too easy to turn enemies into friends on a dime. He suggests house rules that require a dedicated, 10-minute-plus time period to really make NPCs change their minds. He also suggests making changes more gradual, to stop uncharacteristic 180-degree reversals. And he makes a great design point -- any skill that player characters can use on NPCs but not vice versa is inherently broken.
Mearls is not the only one unhappy with diplomacy. In 2003, Rich Burlew put together a variant house rule on handling the skill. Burlew suggests making the diplomacy DC dependent on the target's level, and he adds in a bunch of circumstance modifiers to improve verisimilitude, What's more, while the result of successful diplomacy under the core rules is to make the target friendlier, Burlew instead suggests that the purpose of the skill is to propose a deal and adjudicate the acceptance or rejection of the deal. This allows sensible modifiers like judging the risk of the deal vs. the reward.
People are still talking and talking about Burlew's rules.
Mearls and Burlew both offer improvements to the core rules mechanic, but some gamers may still be unsatisfied. They may look for a social conflict mechanism that is as rich and deep as D&D's mechanism for resolving physical conflicts. There are one or two variant rules I have heard of but not seen for d20 games. This review mentions the one in Mongoose Publishing's Ultimate Games Designer's Companion (d20) (search in the review for "social combat"). And there are evidently social combat rules in the Quintessential Samurai, by the same publisher -- maybe even the same rules.
Here's another d20 system idea I stumbled on at the last minute: instead of rolling a die for a diplomacy check, assign a number from 1 to 20 depending on the actual role-play goes.
And of course, there is a vast wealth of innovative social combat alternatives once you look beyond the d20 system. Here are two online: this Unified Social Combat System discussion and this Cutting Remarks PDF which I found out about at The 20' By 20' Room. Good ideas in both.