Barns@OFFICE@sri-unix (07/23/82)
As one involved in the maintenance of the current Augment system, let me add a few notes on its past and present behavior and rationale: 1) A good short description of the NLS interface is found in R. W. Watson, "User interface design issues for a large interactive system", AFIPS Conference Proceedings, 1976 NCC, Vol 45, pp. 357-364. It isn't up to date, but the discussion of philosophy covers all the issues raised in this Works exchange and is still mostly valid. 2) As that paper explains, the command recognition system was/is wired to allow a multitude of different forms. These are derived by run-time interpretation of a single "grammar" specification. Until recently, the user was allowed to choose any one of 7 modes of keyword recognition. Nowadays we do not allow this because practically everyone used "terse-terse" and it seemed wasteful to document and explain the other modes which hardly anyone cared about. The machinery is still there, under program control. In fact, additional flavors now exist. 3) The interface for direct keyboard (human) use and for process use need not be the same. I contend that an interface which is optimal for one is inherently bad for the other because the bandwidth of the interfaces are so different. Processes (command files, other programs interacting through "pipes", etc.) are relatively fast but stupid sources of input. Humans are slower but (one hopes) more flexible. Augment copes by using "demand" recognition (=what Tenex/Tops Exec uses) when accepting stored commands as input, instead of the "terse-terse" mode. If anyone's interested in hearing more, message me... Bill Barns