1. Unix pipes are 1 dimensional

grep ”^#” dat1 | foobar > dat2

whereas the key thing about this idea is that real pipes can be multidimensional

grep ”^#” dat1 | foobar

| |

| | snafu | stitch > dat2

2. This system treats ">" as a program in itself and verifies a "program descriptor" before it trashes the next one in the pipe

3. Unix pipes give you the error code of the first stopped component in the pipeline whereas in this model you can hook up instrumentation/statistics to any stage of the pipe