Instead of relying on hardware to make the connection between guardrail posts and shallow joists (see Question and Answer, March 2014), I would take a common-sense approach similar to drawing C, which shows the post through-bolted to doubled 2x4 joists. Rather than through-bolting from the sides, though, I would install a 6-inch-long 4x4 block between the joists and behind the 4x4 post, then fasten the post to the block with a single 1/2-inch by 10-inch bolt with fender washers and nuts to lock it into a solid L shape. Then I’d fasten the assembly to the framing with three 3 1/2-inch deck screws on each side of the 2x4s in a staggered pattern, reinforcing the connection with construction adhesive. The drawing also doesn’t show a rim joist, which could also be worked into the equation for added strength. Lots of hardware is not always the answer.
Paul Bennett responds: This solution is similar to one that I contemplated when presented with this problem, and is one that I might have implemented as a carpenter. By using a through-bolt into blocking, you have eliminated the end-grain considerations that might be created by lag-bolting into the 4x4–this is good thinking. But the fundamental problem with your solution—and with the similar one I contemplated—is that the 2x4s will still likely split along their length, making the connection unstable. Although calculations would say otherwise, your detail may work for a time, but given how severely wood dries, cracks, and rots in many climates, I don’t think it would work for long. This is the kind of thing I see leading to injuries as a deck ages. It’s not possible to walk under a rooftop deck like the one referred to in the original question (remember, limited clearance is what prompted the need for shallow joists in the first place) and easily inspect for shrinkage cracks or decay in the joists—especially at the end of the joist with screws in it—that will make this post unstable. Calculations aside, this is reason enough to not do it this way.
Having worked as both a carpenter and an engineer, I can appreciate your comment about hardware. However, there is a reason engineers often default to hardware. By specifying a $30 piece of hardware, we are relying on something that has been tested and has published load values. The $30 spent in the field saves $100s in engineering, since we don’t have to do the math.