Jon,
I share your surprise, however I have long moved on from such surprise that this matter is not covered really at all in Western practice.
In only one western book, a Ranford text on using the steel square from 1906, did I find any reference to this problem, where he shows a simplistic method of backing the post after the timber has had its end cut to meet the floor cleanly.
It's really bizarre to me how little this subject has been covered in Western carpentry despite the inherently very basic arrangement of a wooden structure having splayed legs. So many structures have had this form. Maybe some folks figured it out in the past, but they appear to have kept their approach and solutions to themselves. The French cover it in certain ways, however as often as not they deal with a splayed leg by rotating the leg entirely so as to bring one face into alignment with the prism rather than changing the leg shape. As expected, this results in some joinery issues, and as often as not the connections in that sort of French work are spiked together.
You are correct that with slight slopes the effect is less pronounced. In fact, doing this form of compound joinery work when the leg splay is slight is more challenging than otherwise. Still, geometry is geometry, and if the leg shape is not adjusted then the joinery problems arise elsewhere.
Both articles linked, reveals how the geometrical issues are usually avoided - - the joinery employs floating tenons for the stretchers rather than through tenons, or the stretchers only have stub tenons. These are weaker, glue-dependent connections it should be noted.
The Garrett Hack article annoyed me when it first appears and continues to annoy me on the re-read. Yes, he doesn't seem to mention cutting the top/bottom stretcher abutments at a slight angle. I decided to do a quick Sketchup model myself to see how close he would be, given the 4˚ splay angle mentioned in the article:
- 4˚ splay.jpg (41.85 KiB) Viewed 7496 times
As you can see, on the top/bottom stretcher surfaces, a 90˚ cut would be out by 0.14˚, And the apparent error on the broad faces of the stretcher is negligible - likely a sketch Up error.
I also drew the stretcher lower down, without the upper surface beveled to match the top of the leg to confirm that the 90˚ angle would still be out (should be the same as the above drawing), as he processes the cuts with the stretcher as a rectangular section:
- 4˚ splay 2.jpg (45.79 KiB) Viewed 7496 times
Yup.
My guess is that he assumed a slight cut out error and simply made a little adjustment to the tenon shoulders after he tried the fit. Or maybe he just clamped it up tight and kind of squished the grain down on the leg. Maybe the cock-beading covers it up a bit too. I don't know, but it is articles like that which have led me to not bother with that magazine any more.
If the leg shape is not backed, then the stretcher top/bottom edge bevels are obviously going to be different, however the connection can certainly be made cleanly - if Hack had taken any time to actually explore the work he is tackling to a more detailed level, he could have cut the stretcher end cuts to the correct required angle to meet an un-backed leg cleanly.
As soon as you have a through tenon however, then the issues of centering the tenon in layout would crop up and surely be noticed, though who know, maybe i overestimate what people notice or don't notice? If the connection is blind, this compound reality just passes without notice.