An old paper on boxer that I found while reviewing related work for a paper I'm writing, abstract:
Boxer was designed as a successor to Logo, with the same educational goals in mind. Whereas Logo has incrementally added features over the years, Boxer changes the
core computational structures of the language and environment. The aim is to make
learning easier and more rewarding, especially over the long term.
As an early graphical language, it is quite interesting. It is one of the first works I know of that dive into concreteness (they call naive realism):
Here, the idea is that the user of a computer system can pretend that what appears on the screen is the computer system. That is, you donâ€™t need to do a lot of mental work interpreting an abstract presentation that relates only indirectly to the fact of the matter (as, for example, imagining something called a variable that is changed or accessed by commands). Instead, naive realism means that everything in the system must have a visual presentation that allows easy interaction with it, including creating it, changing it, moving it, and deleting it.
As well as one of the first languages to use spatial relationships in programming (the only other I know being...agent sheets):
Boxer uses space and spatial relations systematically to represent aspects of â€œabstractâ€ computation. In particular, Boxer has a wonderfully transparent hierarchical structure of boxes inside of boxes that represents huge ranges of computational meanings.
But what I really like is this notion of pokability:
Logo has always, unfortunately, distinguished in one way or another the mode of creating procedures from the mode of executing them...you cannot easilyâ€”or at all!â€”see the effects of a procedure at the same time that you look at its form. This makes learning by inspecting difficult; you have to flip back and forth between different areas to see a procedure and its effects. In addition, it rules out a mode of learning by interacting with pieces of code, which is very powerful and characteristic of Boxer. For example, if you look at a line of code in Boxer and wonder what it will do, you can just double-click on that line, and it will be executed. This also turns out to be an extremely powerful debugging technique. If something goes wrong, you can just step through the process by executing one line at a time. That is, the inherent inspectabiliy of Boxer is extended with â€œpokability.â€
And so on....