This article plainly states that a grounding conductor must be run with the supply conductors to a detached structure. Then reading on down to (2) it starts out "Where (1) an equipment grounding conductor is not run with the supply to the building or structure, (2) there are no continuous metallic paths .....the grounded conductor run with the supply shall be connected to the disconnecting means...."
So does this give up permission to run 3 conductors to a detached building if it meets 2 & 3? (under the 05 NEC) Or is this referring to existing installations?
Because that's who's going to be making the call as to such design specifics.
On the economics, I can't imagine why any electrician wastes his time trying to eliminate bonding or grounded conductors -- for that is what your query is driving at.
There's no money there.... However, there's a very good possibility that your work will be red-tagged if you omit such conductors.
Today's AHJ realize that it is essential for grounded conductors to be bonded to the GEC at only ONE POINT.
Section 250.32 (B) is all about that.
In case (1) the dependent panels in the out-building are NOT to be without GEC systems -- but their grounded conductors are NOT to be bonded there.
In case (2) the dependent panels in the out-building are REQUIRED to be bonded to their GEC systems since in this case only, the isolation of the out-building permits it to be treated, electrically, as if it was an independent Service -- wiring it accordingly.
Case (2) demands conditions that block circulating currents on neutral conductors/ grounded conductors.
The break in design thinking occurred in 2002's NEC.
Actually it is not for any installation I am planning. What happened was I was at the supply house when I overheard one of the counter guys telling a customer(DIY) he could run a 3 wire feed to a detached garage. The scary part was he told him he could do all the grounding at the garage with ground rods with no mention of bonding the grounded conductor. I butted in and told him he was wrong on both counts. At that point all the counter guys told me the local inspector told them it was ok. The code language on the number of conductors required is not clear to say the least (surprise).
Aside from the number of conductors required, it was obvious they didn't understand the importance of bonding the grounded conductor where no EGC exists. They really believe the ground rods will take care of that. One of them argued that the grounded conductor is already bonded at the main service.
The only thing I will give you to chew on is why do we ground the service at the service disconnect? It is already grounded at the transformer.
Basically, because the origination of the EGC needs to be grounded, and a service drop does not include an EGC.
A non-EGC run to a detached building would originate an EGC, so that must be grounded, too, much like the service drop.
A run to a detached building that includes EGC (4 wire for 120/240 split phase) better protects other metallics (twisted pair, coax, door bell, ethernet ... a list that was not so large just a couple decades ago) between the buildings by not having any current carrying conductors grounded at both ends. The EGC still needs to be grounded (again) at the detached building, right?
So what if we did have separate EGC on a service drop? We'd ground that at the service entrance or disconnect in the served building. What if the MV system neutral connection back to the substation breaks (e.g. lightning or or falling tree branch takes it out)? The return then is ground ... any ground anywhere. By having grounding near the building, shared with incoming metallics, at least we can significantly reduce the step distance experienced at the building, and thus the MV potential being radiated out from the transformer location (even if it's a pad).
At least with the service drop not having its own EGC, any circulating currents have to involve the resistance of ground.
There is a ground wire on every pole I see around here with a transformer and the grounded/grounding conductor goes from pole to pole, getting regrounded each time along the way. There are two ramifications to this. One is the earth has become a current path and Two, for this discussion, the neutral conductor is already grounded when it comes to your house. We reground it there because we want to create a local ground reference at the house. That is also why even the old version of 250-24 or 250.32 (depending on how far back you go) required the ground electrode at any subsequent building.
So far, everything would remain consistent if we stayed with the 3 wire feeder wouldn't it?
Each building would have it's own "service" and be treated the same. We even have it in the code that each building shall only have one source of power with few exceptions, disconnects etc (225 and 230 look virtually identical in this) so ground shift and voltage drop issues on the grounded conductor are mitigated by the GES and the bonding jumper. If I have a well house, out in the yard, with a little panel serving the pump, a light and maybe a heater, where is the hazard?
<now I turn my hat around>
The problem came up when we started stringing communication cables between these buildings and creating alternate paths. In most cases it is just a hazard to equipment but it can create a personal hazard. The first language, about 20 years ago, to deal with this was the "no continuous metallic paths..." limitation on the use of a 3 wire feeder. Unfortunately a path that did not exist at the CO may show up a week later. In that regard, the new rule makes sense.