Construction of arbitrary regular polygons with ruler and compass A lot of arithmetic geometry has to do and took off with constructions or proofs of non-constructibility of specific figures only with ruler and compass. See e.g. the non-constructibility of the regular heptagon and the constructibility of the regular heptadecagon by Gauss.
I wonder what it means that there is a simple device that can be constructed with the help of a ruler and a compass that together with ruler and compass allows to construct arbitrary regular polygons very easily.

The device is nothing but a solid cone of finite height and a flexible
  string of length smaller than the circumference of the cone.

Now arrange $n$ dots on the string at equal distances and close the string such that also the first and the last dot have the same distance. I.e. you arrange $n$ dots equally on a closed string. This can be achieved with a compass alone.
After that you place the string on the cone, keeping all dots at same height until they all touch the cone. Projecting them vertically on the plane gives the corners of a regular $n$-gon.
Note that the cone is not a magic device like the angle trisector which may help to construct a rectangular heptagon, but adds nothing genuinely new to ruler and compass (because it can be constructed with the help of these). On the other hand, a flexible string is enough to create a ruler (as a physical device), and you need two rulers to make a compass. So a string might be enough for everything, in a specific sense? Note further, that angle trisectors - and more generally angle $k$-sectors - can easily be constructed for any angle $\alpha = 2\pi / n$: just place $kn$ dots on the string.
My question is: 

Is there something wrong with my device? Might it be not constructible
  with the help of ruler and compass alone? Do other parts of the
  described procedure add something else beyond ruler and compass (the
  cone itself is supposed not to)?

I guess, it's leaving the plane that "disqualifies" this kind of construction. Might this have to do with leaving the real number line and entering the complex plane when solving polynomial equations? ("Suddenly all polynomial equations can be solved!") On the other hand: The compass is a genuinely three-dimensional device, too!
[Counter-historical note: If geometers would not have restricted themselves (deliberately?) to ruler and compass, there would have been no necessity to study the constructibilty of regular $n$-gons - and lots of arithmetic might not have been discovered (or only considerably later). On the other side: Wouldn't complex numbers have been discovered earlier, when the cone device had been considered more seriously?]
 A: The basic misunderstanding here is that "ruler and compass" is not really about which physical tools you're allowed to use.
When this phrase appears in mathematics it is a shorthand reference to the idea of basing constructions on the three first postulates of Book I of Euclid's Elements:


*

*To draw a straight line from any point to any point.


*To produce a finite straight line continuously in a straight line.


*To describe a circle with any center and radius.

(The fourth and fifth postulates are not basic constructions but are claims about what happens when certain other combinations of constructions are performed).
These three postulates are the definition of "ruler and compass". Often in English the alternative phrase "compass and straightedge" is preferred, to underscore the fact that there is none of the basic allowed constructions that depend depend on having measuring marks on your ruler.
In fact even without leaving the two-dimensional paper there are things you can imagine doing with a ruler that has measuring marks on it -- so-called neusis constructions. These are nevertheless not part of the basic operations. They can, for example, be used to trisect angles.
Mathematics is not about physical tools. Saying "ruler and compass" is merely supposed to remind you what the fundamental three constructions in Euclid are. You may, if you wish, object that it is a confusing or misleading shorthand -- that is ultimately a subjective judgment -- but this does not change the fact that what we really care about is not the tools named by the shorthand, but the three postulates it attempts to point to.
A: That's actually very clever! It shows that the ruler and compass model may actually be narrow-minded, since flexible strings are available, as is paper.
However, I believe that if flexible things are not allowed, then we still have unconstructability proofs. For instance, suppose everything we cut out is solid. Then we are only allowed to place it on the plane, and everything we get will have been constructed in the plane before cutting out the shape. So only if we are allowed to bend and stretch things, the third dimension gives additional possibilities.
