When the players ask for an appeal to review, the first decision the TV Umpire should make, before any technology beyond the action replay becomes available, is whether or not the on-field Umpire was obviously wrong, or whether it was marginal. A marginal decision should stand, even if Hawkeye shows otherwise. The current version of the marginal decision for LBW (when the ball is shown to graze the stumps) is too narrow. The TV Umpire has to have greater discretion.
While the Ashes Test at the Woollongabba exposed the absurdity in the economy of referrals set up by UDRS, three (and i will extend these to four) decisions at Durban and the response to them have brought the issue of the marginal decision to the fore. It is an underestimated issue, because especially in the case of the LBW, it changes the game profoundly. It is not as though the interpretation of the LBW law has not changed at all over the course of recent decades. The interpretation of the law has been different across geographies as well. For example, an entire generation of Indian batsmen were not good players of the sweep shot, because Umpires in India would give LBWs anytime a batsman missed a sweepshot in front of the stumps. This was a special case of the more general rule, that a batsman got very little benefit of doubt if he was playing across the line in front of the stumps. There was a period in cricket when bowlers found it nearly impossible to get LBWs when the batsman was playing forward. The reasoning here was that the ball had too far to travel (somewhere in the range of 6-8 feet from the point of impact) when the batsman was forward. Recently, there has been a period of distinctly more generous LBW decisions (from the bowlers point of view). My theory about this is that the re-emergence of spin bowling due to Warne, Kumble and Muralitharan from 1990 onwards has caused this shift.
Hawkeye and UDRS impose a highly technocratic intepretation of the LBW law, based on a newly manufactured certainty of the Hawkeye prediction of the path of the ball. This prediction has not been subjected to as much scrutiny as it should be, and this is because the ICC is only indirectly responsible for Hawkeye, which remains mainly within the purview of the broadcasters and the companies producing the technology. All these details are swept aside, because of the apparent certainty that the checklist based intepretation of the LBW law provides.
Umpires adjudicate a contest between batsman and bowler, and especially in the case of the LBW, this aspect of the law is important. For example, if a batsman is padding up to a spin bowler as a tactic, then the Umpire is prone to become less and less sympathetic to the batsman's case. This is how it should be. Umpires also adjudicate for bowlers. How many times have we seen Anil Kumble or Shane Warne appeal for LBWs from their sliders which were not given? The bowler knows what he has delivered - there are two reasons why he may not get the LBW - the Umpire isn't reading him, just like the batsman, or the Umpire reads him, but isn't convinced that the ball would have gone on to hit the stumps.
The three LBWs at Durban which went against South Africa would have benefited from the UDRS, provided an initial subjective judgment was made by the TV Umpire about their marginality. In such an event, I think South Africa would have gotten the Ishant Sharma LBW given, provided hotspot showed no inside edge. The deVilliers LBW would have been ruled to be marginal and the on-field decision would have stood provided the TV Umpire was convinced that the ball didn't hit the glove first (the general consensus seems to be that it didn't, I agree with this). I am less certain about what the outcome of the Boucher LBW would have been. Contrary to the general view, I think that ball did come in a little bit. It was a bad leave, and like deVilliers, Boucher was not offering a shot. If I was the Umpire (a totally amateur one), I think I would have asked the on-field Umpire to reverse the decision, because even though Boucher deserved very little benefit of doubt given that he wasn't offering a shot, it didn't look at good on replay as it did in real time, when the fact that Boucher was leaving it played a larger role (and perhaps much larger than was merited) in the decision making.
I would add a 4th interesting LBW decision to this - Hashim Amla's first innings LBW. It was a marginal decision in my view, which of course is viewed very favorably by UDRS afficionados, only because the little green arc of Hawkeye showed the ball heading towards the leg bail. Amla was sweeping, was hit on off-stump, and the ball was turning and bouncing. It was a close decision, which could have easily gone against Harbhajan Singh. I'm sure India would have complained, but it would still have been marginal. The marginal decisions have gone against South Africa in at Durban. Of this I have no doubt.
UDRS, if it must be applied would be far better if the TV Umpire is allowed to first make a subjective judgment, based on just the replay about the marginality of a decision. If the TV Umpire thinks a decision is marginal, and lets the on-field decision stand, then the side asking for the review should not lose one of it's reviews. Unless this is done, UDRS will fail in it's focus on curbing obvious errors. The effort to eliminate subjectivity in Umpiring decisions is futile. Furthermore, in the process, interpretations of laws have shifted imperceptibly. It is this imperceptibility that is the problem.