Sounds like an interesting application.
I guess the first question is what is the required timing accuracy to measure the speed of the skier to some level of accuracy.
For this purpose Ill ignore the errors introduced by the uncertainty of the actual sensor measurement. In real life you might have to consider that the posture of the target changes between measurements thus changing the measurement point on his/her person. Also you may need to filter out glitches caused by things like the ski pole blocking the sensor ahead of the main body. There are probably lots of things you can do to reduce these effect. E.g. by averaging out the sensor ‘entry and ‘exit times.
Ill make some assumptions
Max ski speed = 50km/h = 13.88m/s
Gate distance = 10m
The basic formulas are
Time = Distance / Speed,
Speed = Distance / Time,
Distance = Speed * Time
Time between gates then = 10 / 13.88 = 0.720
What time would give us a speed of 51km/h(14.16m/s)
Time = 10 / 14.16 = 0.706.
So to be accurate to less than 1km/h for a skier going ~50km/h your timing measurement must be accurate to less than 0.724 – 0.706 = 14milliseconds.
Based on that. If the interrupt handler works for IR as Cowboy says I would guess that it should be accurate enough for your application.
My first step would be to hook up an accurate signal generator (or use the 1PPS from a GPS module) and test the accuracy you can achieve.
The only other way I could think of is to use the SPI interface to clock in the raw digital inputs. You would probably have to put some simple external circuitry to switch from one sensor to the other, but you could achieve sub millisecond timing this way. I have seen someone do this on a dsPIC to sample a signal at over 1MHz.