The old type switch is a fairly complex item with nine parts:
a top, combining two curved rails,
a bottom cover, glued to the top to protect and hide the parts inside,
a point, from red plastic,
a holding plate, not visible,
a spring to load the point,
a rod to hold the spring,
a rotating lever, from yellow plastic,
a spring to hold the rotating lever in one of its two positions
a cap for the lever, to rotate it more easily and indicate its state.
I opened the switch by drilling out the knobs on the bottom cover plate with which that is glued into the top. First I tried to pry the bottom off, in the hope that it might “click” off, but it is firmly glued. After getting nowhere but damaging the cover, I finally drilled away the glued buttons with a 5mm bit (see last image).
A switch has three tracks connected to it, which I will respectively call the forking side and the two joining sides.
A train can go over a switch in two very different ways: if it comes from the track connected to the forking side it can be guided to either one of the two joining tracks, depending on the position of the point.
A train coming from either of the two joining tracks has no choice: it will have to go out over the only track connected to the forking side.
Operation and Memory
The rotating lever has only two stable positions: guiding the trains coming from the forking side to the left or guiding them to the right. A transparent plastic rod transmits the position to a spring that then pushes the point into its position.
A train coming from one of the joining sides may encounter the point set in the right position for the joining track it is on, but it may also find that the point is set to the other joining track. Because the point is positioned by a spring, the train can push it out of the way if necessary, and the point will then spring back to its original position when the train has passed. Thus the old style switch has "memory" in that it remembers how it was set, despite perhaps having been pushed the other way.
Switches: New Type
The new switches have only two parts: rails and point. A train passing over a switch from a joining track with the point set to the other joining track will push the point to the other state, thus "erasing" any previous decision. This may be important if we want to make a layout that performs some computation. In addition, since there is no state spring it is easy to leave the new switch in between guiding left and guiding right.
Real switches have points that cannot be moved by the train, and passing over a point from a joining track with the point set to the other joining track will often lead to derailment of the train. All popular model railways I know of have spring-loaded points to avoid derailment.
The difference is in the traction ridges: in variant 1, presumably the older one, the ridges correctly stop on the right side when they begin on the left side. All curves rails have ridges on the outer side, but not on both sides as one of the wheels has to slip slightly in a curve. The ridges in variant 2 are more like those of the new switch version. Presumably having ridges on both sides for a short length is what ensures enough traction when the locomotive goes over a switch where it encounters other friction from the point. Or maybe it is just an oversight, who knows…