I was watching some videos the other day of factories in Sialkot, Pakistan, and I was especially interested in the vises used in one factory. It’s common to turn vise screws with a sliding round rod permanently stuck through a round hole at one end of the vise screw, but these guys just had the round hole; to turn the screws they had a portable long L-shaped round-rod handle with a right-angle hook on the end of it. By passing the hook through the hole in the vise screw, they had a variable-angle lever available; without removing the hook from the vise, the handle could then be rotated to be parallel to the vise screw, perpendicular to it, or anywhere in between, thus providing them with whatever mechanical advantage they wanted, up to the limit of the handle’s length.
In file swashplate-screwdriver.md in Derctuo, I described a way to
achieve the same thing with a mechanism inside a screwdriver. But I
thought this idea of drilling a hole through screw shafts to turn
them, rather than sawing a slot in the head, was appealing.
Unfortunately, I don’t think it’s compatible with screw heads being
driven in flush with a surface.
The Robertson square drive, Allen hex drive, and Torx drive approaches are also appealing, and although they don’t inherently offer variable leverage, with ball-head screwdrivers they do offer the possibility of off-axis driving, like a sort of universal joint.
With such hole-based systems, there’s a three-way tradeoff between screw-head strength, screwdriver strength, and screw-head size. Making the screwdriver head larger without changing the screw-head size makes the walls of the screw thinner, thus reducing their strength, at least when they are not fully counterbored into equally hard material. Ultimately the strength of the screw against torsion is limited by its material and diameter — enough torsional load, and the head will twist off — but, in grub-screw-like cases where the screw head is not larger than the rest of the screw, I think the weak point is either the screw head itself or the screwdriver shaft where it enters the screw head, because these both must have scantier cross-sections than the solid neck.
I wonder if you can get some advantage by making a very deep hole drive, perhaps going all the way to the end of the screw. This would unavoidably weaken the screw, though. If you can’t, is there a natural limit on how deep you can go before you stop gaining an advantage?
In hole-based screw heads like Torx, you could make the screwdriver head slightly helical in order to pull the driver down into the screw head as you’re twisting it, by way of elastic deformation of the screw head. This would work best if you had separate screwdriver and screwundriver heads with helices in opposite directions, so that an increase in available force in one direction doesn’t translate to a decrease in the other direction.
By making the hole in the outermost part of the screw head circular and gradually tapering it into the shape of the square, Star of David, etc., you could make the screwdriver auto-align to the screw head’s rotational position as it is inserted, at least if the user is providing enough torsional compliance during insertion. This would be very convenient, especially for screws you can’t see. Similarly, you could taper the hole outward to the full diameter of the screw head so that the screwdriver auto-aligns to the screw head’s translational position as it is inserted, as box-end wrenches have done (on the wrench side) for many generations.
Probably a buttress thread profile like the ones used on toothpaste tubes would be a better default (for fasteners, maybe vises and the like too, though they might be better off with planetary roller screws). I think it roughly doubles the screw thread’s strength against pullout.
All in all, though, I think screw fasteners were a mistake. Much of the above also applies to camlocks, scroll drives, etc.