The aircraft transponder system operation would be difficult to implement for rail and wouldn't provide the same degree of protection anyway...
Then, once you do all that, what does it get you?
...
You can't just take a system designed for another transportation mode and apply it where it wasn't intended, and expect it to be useful.
Well, in my defense I already spent about half of my original post about this explaining why it isn't any good
So I'm not saying the freight railroads should do something like this at all. But for sport's sake:
1) I acknowledge the line-of-sight thing is a great plus for aircraft and makes it work really well over long ranges, and that buying expensive equipment to make this happen is impractical. Fortunately these transponder interrogation and reply signals would be tiny information-wise and require very little BW. So you have two choices: integrate it with voice is the best one. You could just throw a little signal right on the same voice channel outside of audio range and if you design everything right have the repeaters that are in place handle it, and it just works without buying any new equipment. Choice two is to find a secure way to do it on a lower frequency so you will have some sky wave propagation and range is no longer a problem.
2) As I said originally, you need a way for the train to know what track it's on (and preferrably speed and forward/backward, more later). It sends this information in response to interrogation. Same as you need aircraft transponders to reply with altitude information for air traffic control radar to work with the best accuracy. I'm not sure of the best way. It would help if I knew more about train signaling technology.
3) This is only for collision avoidance! I'm not proposing at all to make a whole new top-down train dispatching system based on transponders (surely GPS would be a better way to do this). So I'm not worried about any data links, computer screens, or gathering the information in a central location at all. The only entities communicating in this scheme is one locomotive to all others within range. Furthermore, since it's a 1-D problem, we can get away from having to know bearing and angle of elevation like for airplanes: distance is enough! If my computer continuously knows the distance between Alice's train and Bob's train (and of course what tracks we're on), that's enough to predict a collision. Now, we still need to know if that's distance in front or behind to tell if we're about to run into Bob or if he's overtaking us from behind! If we send the current "speedometer" reading along with info about what track we're on when our transponder is interrogated, we need only absolute distance, and can do away with directional antennas and all the rest of that!
Notice the analogy to TCAS, which mediates a discussion only between two aircraft in order to prevent a collision, and does not communicate to any centralized "dispatch".
One way around the "which track" problem is by having a different "squawk" code, for each track, siding, etc. Now if Alice and Bob are approaching each other on a double-tracked mainline on different tracks, their transponder units ignore each other because they're using different codes. One could envision maps of the route showing where the engineer must change squawk codes at certain mileposts as the train changes tracks, similar to how such a map would show slowzones today. You could find ways to automate this...
So once you put all these refinements in place that I have described above, I think you have a collision avoidance system that would work just fine. It's got some more subtle potential problems left that I won't get into. And with relatively old technology and little capital investment beyond an antenna and radio added to each locomotive. Just saying it is feasible.
(Anyone want to start a company with me? HA! :lol: )