I'll admit, detail B is easier to construct if I'm putting my contractor's hat on. Detail A just gives me the warm and fuzzies regarding rotational stiffness of the connection.
I'm not seeing the curve. Is it curved in the plane of the roof? Because that may dictate the answer here. If you rotate the stub to be more-or-less in line with the beams such that you can fully weld around the tube and develop a moment connection through there to remove the pin, I like B a lot more. The more extreme the angle of the beam to the wall, the better this one will likely perform (you'll eventually run out of weld access room to the plate on one side of the tube as the angle changes).
Ben29:
Why doesn’t the bent HSS 8x6x.25 run continuous over the wall, with a bolted (?) connection at the ridge? Then, you just need a sloped seat connection on the wall; a base pl., a piece of 8x6x.25 tube, slope cut and welded to the curved beam, welded to the base pl., and with two side pls. approx. 6" wide by 8 or 10" high by .25" thk. with fillet welds to the two sections of HSS 8x6.
Nothing amazing here. I just don't have it in me to not submit an entry to a good steel detailing challenge. If I understand dhengr's suggestion correction correctly, it's similar but with field welded gusset plates on either side of the beam to beam connection instead of the seat angle and bolts. Mine is not a connection that would be great for moving a ton of torsion so, if there's plan curvature without roll bracing, this may not be the way. dhengr's connection would do better under significant torsion.
It appears that this splice is at a point of inflection (assuming curvature is ONLY in a vertical plane) - which if true means each piece has single curvature bending, which should be [citation needed] a net fabrication savings.
If that's the case, I'm partial to detail B. Not a huge fan of through-bolting if I can avoid it.
Please note that is a "v" (as in Violin) not a "y".
