Yes, that old argument. This discussion on LinkedIn is interesting, but will not die. Someone asked a good question today, and it led me to an interesting place.
I am basically putting it up here so I can have it preserved somewhere for my own purposes.
Philosophically, I tend to fall back on precepts of, say, Good Design as it applies to furniture and so on. We must understand our medium, so we can put it to the best use. Even more so, we should allow the nature of the specific material (for us: function, widget maybe) to be self-evident, and not twist it to something else or obscure it's true function.
This is impossible to do when the nature of the medium is a sketchpad, or Fireworks. Not to take an analogy too far, but consider furniture design. Furniture designers do not work in furniture factories. Ah, you may say, but they can make one-off pieces, and prototypes, using the same basic materials and techniques; one piece of wood is cut very much the same way by hand as it is when mass produced.
I say there are degrees here, but more importantly, look at newer technologies. There is no workshop-scale analogue to rotational-molding. The Bubble Club chair, say, never /existed/ before the first one came off the assembly line. It was prototyped, or sketched (either analogy works) in cardboard if memory serves, maybe with some clay, and literally sketched on paper, diagrammed and then coordinated with manufacturing to implement it.
Much the same exists for any number of other design fields. Industrial designers, in general, sketch from foam and clay and plywood, what they want to be molded in plastic and aluminum and machined from steel. But they also know perfectly well the material properties, and choose the right ones for a particular application, work with design engineers to make sure it will work right, electrical engineers to make sure there is enough room for the control panel, manufacturing engineers to make sure it can be built as expected, cheaply and quickly. Etc.
Why can't we, as Interaction Designers (etc for other titles) come to terms with both specialization and collaboration, and work the same way? Not to mention my gripes with the field which is the most easy-to-update product ever (app or web) being so terribly inept at considering maintenance, and revisions, and efficiency through reuse.