Before I get covered in detritus cleaning out the milling head on my Bridgeport, prior to replacing the broken drive belt here's my summary on the design and manufacture of bevel gears. It's based on my article in the current issue of MEW (216). Was that a trumpet I heard? 
Bob is correct in that for a straight tooth bevel gear the tooth size changes from a maximum at the outer edge to a point where the two axes intersect. The shape does not change, only the size.
Bevel gears are normally designed using the diametric pitch at other outer face.
Unlike a spur gear the involute tooth form for a bevel gear is not directly related to the number of teeth in the gear, but to the number of teeth divided by the cosine of the pitch cone – Tregold's approximation.
Bob is also correct that a true bevel gear cannot be cut on a milling machine. However they can be cut on a shaper, with the aid of a Bilgram bevel gear generator.
There are two methods for cutting an approximation to a bevel gear on a milling machine using involute cutters. Both methods use three passes, one through the middle of the space between teeth and two with the blank offset and rotated so that the outer space can be enlarged without affecting the inner edge space. For smaller diameter gears thw first pass can be omitted.
The first method takes the standard design of bevel gear, ie, using the DP at the outer edge. An involute cutter of the appropriate DP and number of teeth is used based on the calculations at the outer edge. However, these are not standard involute cutters, they are narrower than standard, so that while they have the curve associated with the outer tooth shape they are narrow enough to fit through the gap at the small end. These cutters are normally stamped 'BEVEL'. They do not seem to be available now. This method results in teeth that do not have enough curvature at the small end, requiring the use of a file. I have made bevel gears using this method, but I did have the resources of the main Royal Aircraft Establishment machine shop behind me, so was able to borrow the required cutter.
The second method is called the parallel depth method. Here the depth of the tooth is constant, not varying as in a true bevel gear. There is one other crucial difference, the design is done using the diametric pitch at the small end, so a standard involute cutter can be used. This is fine, except that it can be problematic if the gears are to be used in an existing design where the outer diameter will almost certainly be the controlling dimension. Nice round numbers at the inner edge are unlikely to result in nice round numbers at the outer edge. 
I cut the gears for my traction engine differentials on a 4-axis CNC mill, so they are true bevel gears. However this method will not work with the bevel gears needed for the governor, as the required ballnose cutter will be too small. I have looked at replacing the gears with those designed using the parallel depth method, but it's a PITA due to the afore mentioned calculation issues. I plan to buy a copy of Gearotic and use the CNC mill to machine the governor bevel gears, but with the crucial difference that the rotary table will be tilted over, thus obviating the need for a ball nose cutter.
All this information is in the book by Ivan Law if you read it carefully.
Regards,
Andrew