Without some sort of guidance a high altitude Glider could simply drift off in any direction, so it is necessary to have at least one axis of control (yaw). One approach would be to adjust the rudder and use the feedback from the GPS heading to guide the glider. However, it would be desirable to be able to increase or decrease the airspeed to compensate for wind, especially when travelling through the jetstream. One possible approach would be to use a digital compass like :
http://www.milinst.com/Adons/adontext.htm#sensors
From a magnetic field vector we can only get two euler angles, so we would have to call roll=0, which would be fairly accurate in a stable, slow turning glider if we only wanted pitch and yaw to within a few degrees for flight stabalization (GPS would be used for the actual guidance). The following spreadsheet demonstrates the concept (although it hasn't yet been fully checked) Offset angle and incidence angle are set for UK conditions, and x_one, y_one are relative readings from the two axes of the compass module. The modulus of (x_one,y_one) has to be less than one, and some vectors give #Nan as they have imaginary solutions with roll=0, so any microcontroller on gumstick implimentation of this would have to have a sanity check
yaw-pitch.xls (excel spreadsheet)

Model Aircraft Controls