Why can't you do this?:
1) Build cesium vapor based laser pulse accelerator as described in the Princeton experiment cited above(
http://www.cnn.com/2000/TECH/space/07/20/speed.of.light.ap/)
2) Determine exactly how much earlier a message/pulse is received than it is sent through the cesium.
3) Set up a
buffer for storing messages received from the future (by sending contents of the buffer through the cesium at that future time).
4) Schedule dates for playback/reception start of messages from the future.
5) Start receiving the message. If the time it goes back (due to limited size of the vapor chamber) is only say a microsecond, that means that at least once a microsecond we play the message back (to ensure the reception which already occured).
Example:
Time offset | Action
0.0000000 | Receive "blah" from transmission E
0.0000007 | Receive "blah" from transmission D
0.0000010 | Send "blah" (transmission E)
0.0000014 | Receive "blah" from transmission C
0.0000017 | Send "blah" (transmission D)
0.0000021 | Receive "blah" from transmission B
0.0000024 | Send "blah" (transmission C)
0.0000028 | Receive "blah" from transmission A
0.0000031 | Send "blah" (transmission B)
0.0000038 | Send "blah" (transmission A)
Obviously for practical applications since we're talking about moving data back only a small fraction of a second each iteration, there would be many more than 5 resends, but this is just to illustrate the idea. (with infinite speed if the cesium chamber was a meter across that would be only 3.33ns the message gets sent back through time each iteration)
There's clearly an issue of not being able to send a whole lot of data if it has to be repeated every few ns at best (meaning the transmission must fit in a few ns), but I don't see why it can't be used to transmit data in priniciple.