danielkinguk
commented
1 month ago We could remove the term 'Intent' and replace 'Control' with 'Output,' leaving just 'Input' and 'Output' for the interfaces in Figure 2. This approach works well since it reflects a conceptual architecture rather than an implementation-specific design. This naturally leads into Figure 3, which takes the NDT concept and presents an applied example combined with a Controller, where 'Input' now represents an Intent and 'Output' functions as a control mechanism. Other examples of applied NDT might include planning and sensitivity analyses, where the input and output interfaces may differ again.
billwuqin
boucadair
cfjanz
cheneyzhoucheng
All – I would like to raise an issue. As far as I am concerned, there is only one major problem with the latest draft – there is nothing else from my perspective that isn’t adequately characterized as minor wordsmithing as part of a final editorial pass.
That major problem is the continued presence of Figure 2. This figure was the primary source of the issues expressed by those I represented in feedback. It has to go.
Please let me restate the issue. The first problem generated by Figure 2 is that it tells the world that an NDT is a controller. The second, connected, problem is it conflicts with the other NDT use cases. And a third serious problem is that the Figure was incomplete – a lot of functionality needed to make a controller, beyond the highlighted data and modeling functions, was missing from the picture and not even mentioned in the text (e.g. the closed loops – inner and outer, intent translation, etc.)
The point of the feedback is that it is better to consider an intent-driven controller as a use case of an NDT. Seeing it that way resolves all the problems. The data, modeling functions, and the functional management of both, are what are common to every NDT use case - and also the totality of what is now described in section 7. So call that the NDT. In any given use case, you have to combine these functions – the NDT - with a use case-dependent set of other functions. That is what section 8 addresses: and Figure 3 – which was intended to replace Figure 2 - illustrates this for the case of an intent-driven controller.
This gets you out of having to think about a bunch of different, use case-dependent types or flavors of NDTs.
If we can agree on this change I think we can wrap up quickly.