Tom
I'll try to explain my thinking using the example of a typical SLR built-in flash (GN = 12 in meters @100ISO) which I will call BI and my Vivitar 283 (GN ~ 35) which I will call 283
Scenario 1, BI knows nothing before the preflash, ie there is no focusing information used in the calculation of exposure (might be wrong there, and my entire arguement falls over). Say we are at 3m from the subject, then the preflash is (presumable) very brief, say 1/10,000 sec and it suggests the exposure is f4 (GN/distance = f stop, 12/3=4). It then fires for a typical 1/1,000 sec to give the exposure.
Scenario 2, BI is completely covered with a shroud containing a photo-electric detector (PE) which is connected to the radio transmitter and the radio receiver is integrated into the circuit of the 283. The BI fires its preflash into the shroud containing the PE which triggers the transmitter. 283 is also fired by the receiver for the same(ish) 1/10,000 second pre-flash, but it is a lot stronger, so the amount of light registering on the SLR sensor is also greater and the camera now thinks that the exposure should be f11 (35/3=f12, close enough to f11). When the BI fires into the shroud for the second time (for 1/1,000 sec) to take the exposure, the 283 fires as well (for 1/1000 sec) and you get a properly exposed picture at f12, ie TTL exposure control for the venerable 283

. Of course, if there were two 283s slaved for both the pre-flash and the actual flash, the exposure would be calculated and then taken at f16.
What it really depends on is getting the 283 to turn on and then off perhaps no later than 1/50,000 second after the BI turns on and off.
I hope that makes it clearer. I think my Homer algorithm will work but my electronics aren't up to the task, so I welcome this forum and I welcome feedback as well.
Cheers