In UNIX, everything is a file, or so the story says.  A "file" is anything that
tries to act like a stream of bytes, and obviously this can include much, much
more than data stored on disk.

Having everything (data, hardware devices, network sockets, directories, and
more) as a file allows for great flexibility.  Code written with one type of
file in mind (typically disk data) can be applied to other types of files, even
in ways the original coder didn't plan for.  Most importantly, each program's
input ("stdin") and output ("stdout") are files, and one program's stdout can
be used as another program's stdin, possibly many times in succession,
producing what are known as pipelines.  In this manner, small programs can be
chained together in innumerable possible combinations to produce functionality
potentially far beyond what the programmer initially conceived.

For example, "grep ^r.*t$ /usr/share/dict/words | tr a-z A-Z | sort -r" looks
in /usr/share/dict/words, finds every word beginning with "r" and ending with
"t", and prints it to the screen, capitalized and in descending order.  "cat
/dev/fd0 > ~/img" copies the contents of the floppy disk to a file in the
user's home directory, and "cat ~/img > /dev/fd0" writes the disk image back to
the floppy, which is useful for making duplicates.  These examples make use of
programs that do pattern matching (searching), letter replacement, sorting, and
displaying text, yet when combined they can do very interesting things
seemingly beyond the scope of their original design.