PP (Re: Context-Orientation)

[Raoul Duke]

an aside: I think PP will happen in the not too distant future in the voice interfacing personal digital assistant race among google, amazon, apple, microsoft, samsung. Star Trekkian do what i say type interfaces. (I don't want to use any until there's a GPL one ;-)

[James O Coplien]

Den 22. maj 2017 kl. 15.58 skrev Raoul Duke <rao...@gmail.com>:

an aside: I think PP will happen in the not too distant future in the voice interfacing personal digital assistant race among google, amazon, apple, microsoft, samsung. Star Trekkian do what i say type interfaces. (I don't want to use any until there's a GPL one ;-)

I don’t think this is what Trygve means by PP.

In our AT&T Bell Labs research on customer programmability, we sliced the world roughly into these primary areas:

• Parameterization
• Application-Specific Languages
• Programming

I think what you’re describing is highly contextualized to the domain of application. The lexemes and semantic elements correspond directly to both the end-user mental model and to the intended domain of application. We find much the same in Siri today. It has a very limited vocabulary. There is no correspondence between the ordering of the lexemes (or phonemes) and the ordering of activities carried out by the software. The underlying technology is often rule-based or uses some other non-procedural formalism. DCI as we know it today is restricted to Von Neumann computation. That is not the primary paradigm for most of these systems.

I found an iTunes app that does exactly what Trygve asks for in his wake-me-if-the-weather-is-good example, but it is a piece of software constructed exactly for that purpose (https://itunes.apple.com/us/app/genius-alarm-weather-smart/id901875952). It is an extremely narrow application. The “language” is extremely narrow and has a very limited vocabulary. Each word in a language with a small vocabulary conveys an incredible amount of information. Further, the grammar is further restrictive, which amplifies the contextual information significance of each language element.

There’s probably a reason that Paolo Gui made a very specific app. He could have solved a much more general problem with a broader, more general-purpose language. But he stuck to a very small language. The reason is — as we found at Bell Labs — generality has a high cost. At one extreme is a Turing machine and at the other end is an intelligent, sentient robot. Most programming lies very, very close to the Turing machine, and Baby IDE is no exception. The very best apps, such as the Apple app above, are some ways along the spectrum towards the robot, but the gap to the robot is still very large. The gap is made to appear small with super-contextualization. In the extreme, it looks like this: