Dataflow Shells

[Don Libes]

In article <17062:Feb2120:31:0291 at kramden.acf.nyu.edu> brnstnd at kramden.acf.nyu.edu (Dan Bernstein) writes:
>In article <5503 at awdprime.UUCP> dcheney at dcheney.austin.ibm.com (David Cheney) writes:
>> "the setup of more complicated networks of processes
>> than just pipelines".
>
>I haven't seen any information on 2dsh except for occasional references
>on the net. I have a tool that understands things like
>
>  prog1  3|X  --- prog2 ---  Y|4  prog3
>         X|  prog4 --- prog5  |Y

Some other work along these lines was cited in my Summer 1990 USENIX
paper on expect:

  gsh [3] is based on the Bourne shell, but handles graphs of processes,
  such as sending the output of one process to two processes, or
  building a set of three process in a cycle.  ....

  MTX [8] is a screen-based pipe manager.  It solves the same set of
  problems as gsh, although the interface is mouse-oriented instead of
  keyboard-oriented.  In addition, MTX can rearrange connections in
  use.  It does [this by interposing processes with ptys and hence is
  not as fast as gsh].

  [3] Chris McDonald and Trevor Dix, "Support for Graphs of Processes in
  a Command Interpreter", Software: Practice & Experience, Volume 18
  Number 10, p.  1011-1016, October 1988.

  [8] Stephen Uhler, "MTX - A Shell that Permits Dynamic Rearrangement
  of Process Connections and Windows", Proceedings of the Winter 1990
  USENIX Conference, Washington, D.C., January 22-26, 1990.

Don Libes          libes at cme.nist.gov      ...!uunet!cme-durer!libes