LatLonSpherical.intersection algorithm is different from edwilliams aviation formula.

[aldas]

http://www.edwilliams.org/avform147.htm#Intersection has

IF sin(lon2-lon1)<0
   crs12=acos((sin(lat2)-sin(lat1)*cos(dst12))/(sin(dst12)*cos(lat1)))
   crs21=2.*pi-acos((sin(lat1)-sin(lat2)*cos(dst12))/(sin(dst12)*cos(lat2)))
ELSE
   crs12=2.*pi-acos((sin(lat2)-sin(lat1)*cos(dst12))/(sin(dst12)*cos(lat1)))
   crs21=acos((sin(lat1)-sin(lat2)*cos(dst12))/(sin(dst12)*cos(lat2)))
ENDIF

but same condition seems to be inversed in this library

geodesy/latlon-spherical.js

Lines 426 to 427 in 761587c

	const θ12 = Math.sin(λ2-λ1)>0 ? θa : 2*π-θa;
	const θ21 = Math.sin(λ2-λ1)>0 ? 2*π-θb : θb;

meaning that Williams algorithm for crs12 in case sin(lon2-lon1)<0 is crs12=acos((sin(lat2)-sin(lat1)*cos(dst12))/(sin(dst12)*cos(lat1))) but in this library same condition results θ12 = 2*π-θa;

which one is right?

I am totally clueless at trigonometry but noticed this difference

[chrisveness]

Williams chose to take west longitudes as positive (“For the convenience of North Americans I will take North latitudes and West longitudes as positive and South and East negative. The longitude is the opposite of the usual mathematical convention”).

I have stuck to normal sign convention with west -ve and east +ve.