Just as there are frictional losses associated with moving masses on a surface, what if there are frictional losses associated with moving information on a substrate?  We propose to model these losses as proportional to ``bit-meters,'' \textit{i.e.}, the product of mass of information (\textit{i.e.}, its entropy) and the distance of  information transport.  We use this model to obtain unavoidable limits on energy consumed in "non-frictionless" encoding and decoding of a channel code. These limits are qualitatively stronger than Maxwell's demon-based "frictionless" bounds in thermodynamics of computation, and have implications on the choice of the channel-code that minimizes the total energy-per-bit.