On Tue, May 12, 2009 at 7:16 AM, David Carter <david.m.carter@gmail.com> wrote:
Specifically, a1.txt contains bytes 97 and 10, while a2.txt contains
bytes 224 160 139 237 144 164.

Have I misconstrued things, or is this a bug?

It's probably a bug, and it has (I hope) already been found and fixed. Please try the darcs version of text instead:  http://code.haskell.org/text/