If you compose a Galilean transform (+v) with a Galilean transform (-v) you get a Galilean transform of (0)
If you compose a Lorentz transform (+v) with a Lorentz transform (-v) you get a Lorentz transform of (0)
If you compose a Generalized Galilean transform (+v) with a Generalized Galilean transform (-v) you get a Generalized Galilean transform of (0)
(See
post #21 of "Tutorial: Relativity - what is a reference frame?")
If you compose a MT transform (+v) with a MT transform (-v) you do not get a MT transform of (0) -- In fact, if u and v are
arbitrary non-zero velocities, then no MT transform (u) composed with MT transform (v) equals a MT transform of (0). Thus a MT transform cannot be interpreted as a change of coordinates describing the same physical universe.
--------------------
Length contraction in relativity is part-and-parcel of relativity of simultaneity. Let A & B be objects moving in parallel world lines separated only by a fixed amount in the x coordinate direction.
Then
$$\begin{pmatrix} \rule{0pt}{20ex} 0 & 0 & 0 & 0 \\ \rule{0pt}{20ex} -u_x & 1 & 0 & 0 \\ \rule{0pt}{20ex} -u_y & 0 & 1 & 0 \\ \rule{0pt}{20ex} -u_z & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} t_A \\ \rule{0pt}{20ex} x_A \\ \rule{0pt}{20ex} y_A \\ \rule{0pt}{20ex} z_A \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} x_0 \\ \rule{0pt}{20ex} y_0 \\ \rule{0pt}{20ex} z_0 \end{pmatrix} $$
and
$$\begin{pmatrix} \rule{0pt}{20ex} 0 & 0 & 0 & 0 \\ \rule{0pt}{20ex} -u_x & 1 & 0 & 0 \\ \rule{0pt}{20ex} -u_y & 0 & 1 & 0 \\ \rule{0pt}{20ex} -u_z & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} t_B \\ \rule{0pt}{20ex} x_B \\ \rule{0pt}{20ex} y_B \\ \rule{0pt}{20ex} z_B \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} x_0 - L \\ \rule{0pt}{20ex} y_0 \\ \rule{0pt}{20ex} z_0 \end{pmatrix} $$
Subtracting one from the other, we find:
$$\begin{pmatrix} \rule{0pt}{20ex} 0 & 0 & 0 & 0 \\ \rule{0pt}{20ex} -u_x & 1 & 0 & 0 \\ \rule{0pt}{20ex} -u_y & 0 & 1 & 0 \\ \rule{0pt}{20ex} -u_z & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} t_A - t_B \\ \rule{0pt}{20ex} x_A - x_B\\ \rule{0pt}{20ex} y_A - y_B\\ \rule{0pt}{20ex} z_A - z_B \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} 0 & 0 & 0 & 0 \\ \rule{0pt}{20ex} -u_x & 1 & 0 & 0 \\ \rule{0pt}{20ex} -u_y & 0 & 1 & 0 \\ \rule{0pt}{20ex} -u_z & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \Delta t \\ \rule{0pt}{20ex} \Delta x \\ \rule{0pt}{20ex} \Delta y \\ \rule{0pt}{20ex} \Delta z \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} L \\ \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} 0 \end{pmatrix}$$
so $$\Delta x = L + u_x \Delta t$$ and $$\sqrt{(\Delta x)^2 + (\Delta y)^2 + (\Delta z)^2} = \sqrt{(L + u_x \Delta t)^2 + (u_y^2 + u_z^2 ) (\Delta t)^2 } = \sqrt{L^2+ (u_x^2 + u_y^2 + u_z^2 ) (\Delta t)^2 - 2 L u_x \Delta t}$$.
The point that this emphasizes is that $$\Delta t = 0 \; \Rightarrow \; \Delta x = L \, \wedge \, \sqrt{(\Delta x)^2 + (\Delta y)^2 + (\Delta z)^2} = \left| L \right|$$ or for parallel world-lines, what we call their separation in space
at the same time is constant.
That's a simple consequence of working in Cartesian coordinates for time and space. Now lets look at the generalization of the Galilean transformation for a boost in the x direction.
$$ \begin{pmatrix} \rule{0pt}{20ex} \Delta t' \\ \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} \frac{1}{\sqrt{1 - K v^2}} & \frac{K v}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} \frac{v}{\sqrt{1 - K v^2}} & \frac{1}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 1 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \Delta t \\ \rule{0pt}{20ex} \Delta x \\ \rule{0pt}{20ex} \Delta y \\ \rule{0pt}{20ex} \Delta z \end{pmatrix}$$
Multiplying both sides by an inverse transformation we get
$$\begin{pmatrix} \rule{0pt}{20ex} \frac{1}{\sqrt{1 - K v^2}} & \frac{K v}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} \frac{v}{\sqrt{1 - K v^2}} & \frac{1}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 1 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 0 & 1 \end{pmatrix}^{-1} \begin{pmatrix} \rule{0pt}{20ex} \Delta t' \\ \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} \frac{1}{\sqrt{1 - K v^2}} & \frac{-K v}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} \frac{-v}{\sqrt{1 - K v^2}} & \frac{1}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 1 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \Delta t' \\ \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} \Delta t \\ \rule{0pt}{20ex} \Delta x \\ \rule{0pt}{20ex} \Delta y \\ \rule{0pt}{20ex} \Delta z \end{pmatrix} $$
Substituting into our expression for coordinate differences between parallel world-lines we get:
$$\begin{pmatrix} \rule{0pt}{20ex} 0 & 0 & 0 & 0 \\ \rule{0pt}{20ex} -u_x & 1 & 0 & 0 \\ \rule{0pt}{20ex} -u_y & 0 & 1 & 0 \\ \rule{0pt}{20ex} -u_z & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \frac{1}{\sqrt{1 - K v^2}} & \frac{-K v}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} \frac{-v}{\sqrt{1 - K v^2}} & \frac{1}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 1 & 0 \\ \rule{0pt}{20ex} 0 & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \Delta t' \\ \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} 0 & 0 & 0 & 0 \\ \rule{0pt}{20ex} \frac{-(u_x + v)}{\sqrt{1 - K v^2}} & \frac{1 + K u_x v}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} \frac{-u_y}{\sqrt{1 - K v^2}} & \frac{K u_y v}{\sqrt{1 - K v^2}} & 1 & 0 \\ \rule{0pt}{20ex} \frac{-u_z}{\sqrt{1 - K v^2}} & \frac{K u_z v}{\sqrt{1 - K v^2}} & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \Delta t' \\ \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} L \\ \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} 0 \end{pmatrix}$$
and if we assume $$\Delta t' = 0$$, we have an equation which can be solved:
$$\begin{pmatrix} \rule{0pt}{20ex} \frac{1 + K u_x v}{\sqrt{1 - K v^2}} & 0 & 0 \\ \rule{0pt}{20ex} \frac{K u_y v}{\sqrt{1 - K v^2}} & 1 & 0 \\ \rule{0pt}{20ex} \frac{K u_z v}{\sqrt{1 - K v^2}} & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} L \\ \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} 0 \end{pmatrix}$$
with solution:
$$ \begin{pmatrix} \rule{0pt}{20ex} \Delta x' \\ \rule{0pt}{20ex} \Delta y' \\ \rule{0pt}{20ex} \Delta z' \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} \frac{\sqrt{1 - Kv^2}}{1 + K u_x v} & 0 & 0 \\ \rule{0pt}{20ex} \frac{-K u_y v}{1 + K u_x v} & 1 & 0 \\ \rule{0 pt}{20ex} \frac{- K u_z v}{1 + K u_x v} & 0 & 1 \end{pmatrix} \begin{pmatrix} \rule{0pt}{20ex} L \\ \rule{0pt}{20ex} 0 \\ \rule{0pt}{20ex} 0 \end{pmatrix} = \begin{pmatrix} \rule{0pt}{20ex} \frac{\sqrt{1 - Kv^2}}{1 + K u_x v} L \\ \rule{0pt}{20ex} \frac{-K u_y v}{1 + K u_x v} L \\ \rule{0 pt}{20ex} \frac{- K u_z v}{1 + K u_x v} L \end{pmatrix} $$
In the case $$\Delta t' = 0 \, \wedge \, u_y = u_z = 0$$ we are permitted to simplify even further and write:
$$\Delta x' = \frac{\sqrt{1 - Kv^2}}{1 + K u_x v} L = L'$$
which has two important cases: $$\Delta t' = 0 \, \wedge \, u_y = u_z = 0 \, \wedge \, v = -u_x \; \Rightarrow \; L = \sqrt{1 - K v^2} L' \\ \Delta t' = 0 \, \wedge \, u_x = u_y = u_z = 0 \; \Rightarrow \; L' = \sqrt{1 - K v^2} L$$
In Special Relativity, $$K = c^{\tiny -2}$$, so the separation between the world-lines (L or L' depending on the relation between $$u_x$$ and v) is longer in the coordinate system where A &B don't move.
In Galilean Relativity, $$K = 0$$, so the separation between the world-lines is constant.
In Fan's pet theory, $$K = -c^{\tiny -2}$$, so the separation between the world-lines (L or L' depending on the relation between $$u_x$$ and v) is longer in the coordinate system where A & B move.
But Fan's experiment favors K > 0 over K = 0 or K < 0.