Movement commands should use up more "time" than program instructions

[jorendorff]

Right now each thread only goes 6 instructions per "tick", which is not fast enough; and movement only costs one "instruction", which is too fast.

[bsmedberg]

I'd like to propose:

• computation should be cheap but not free: one tick for each primitive and loop beginning.
• User functions shouldn't cost anything intrinsically, only the primitives they use?
• looking should be a small constant plus additional time the wider the width, i.e. one ticks + an additional tick per 5 degrees of width
• turning and moving should be a small constant plus significant time for the size of the movement. Turning maybe two tick plus two ticks per 5 degrees of width? Movement 2 ticks plus a tick per pixel?
• Shooting should cost some time: maybe 5 ticks?

Thoughts?

A future advanced option might be to customize your turtle to your program, so you can allocate "turtle speed points" so that your turtle moves, shoots, or looks faster.

[jorendorff]

All of this sounds good, except I don't think the cost of looking should depend on the width. If you use a large width, you naturally get less specific information back (look does not tell you the precise direction to whatever it is you see, and you only see the nearest thing in the specified wedge).

User functions and loops are the same thing in the turtle language, which makes item 1 and item 2 a bit tricky to reconcile.

Another thought I had was that it might be neat to have APIs that just mean "start walking", "stop walking", "start turning left", "stop turning". I'm not totally sure that would be more fun. Turtle speed points would be a blast.