Since the LSI-Unix system does not support pipes, the shell has been
modified to simulate pipes through the use of disk files.  When a
command line which requires the use of a pipe is detected, a disk
file is created and opened for reading and writing.  The file is
immediately unlinked, so that the name is available for another pipe
right away.  The file is called '.__pf'.  Hopefully this name will
not conflict with any user file names.
Thus, the symbol '|' in a shell command line becomes equivalent to
\&'> .__pf ; < .__pf'.  The command:

	% foo ^ goo

translates into:

	% foo > .__pf ; goo < .__pf

The process writing on the pipe writes everything into the file, and
when it exits, the reader process is swapped in.  It reads what the
writer has written on the pipe file.  The only danger in this type
of implementation is due to the limitation of space on the floppy
disks.  A very large amount of information going through a pipe could
fill up the disk.

Except for this, the fake pipe code is transparent to the user.