
This topic had already been addressed in 1990 by the IETF in RFC 1149, 'A Standard for the Transmission of IP Datagrams on Avian Carriers'

In all seriousness:
The bird carried 4.75 Mbps, which is approximately equivalent to three T1s at full utilization. The network speed of 190 Kbps is lowish, but within tolerance of residential ADSL upload [which is usually pretty crappy].
Sneakernet is a viable mode of transfer for large data sets, to the point where Amazon Web Services now offers enterprise-grade sneakernet uploads to their data center

- Z
My Thought is if a Bird can be trained to drop of Data it could also be trained to drop it off someplace else first then go to the correct place (aka someone steals all your data and gets it before you do.) The Next question is well why not just steel it. Well if you did that when the bird didn't show up the other person would know, but it shows up around when it should no one would be the wiser.
I heard about this on a radio show and remembered thinking that I would rather have slow data transfer than the possibility of disease transmission by pigeons. Pigeons are somewhat filthy in their general habits.