Moderator sent me a message:
"Making references to other forums are not considered as valid references. Please re-read the PF Rules that you had agreed to".

Ok. If it violates the rules, I am ready to you to explain the incomprehensible. I repeat that this problem has long been solved Котофеич, peregoudov, Пью Чай Ли, В. Войтик.
Given: Two material points, acceleration from a constant proper acceleration from rest in the positive direction of X axis so that the distance between them is relatively laboratory frame (T, X) is always constant and equally. The radius-vector of their relative position is also oriented along Х.
Search: position of the front point on the frame back point (t, x) to the time t in this frame of reference.
Solution. It is essentially based on the transformation of Moller from laboratory frame in the frame associated with the back point.
(1) T=\frac{1+Wx}{W}shWt
X=\frac{(1+Wx)chWt-1}{W}
The calculations, which can be found
http://arxiv.org/abs/physics/9810017
or
http://www.sciteclibrary.ru/rus/catalog/pages/9478.html
argue that the length of rigid rod own size x, the back point, which moves with the acceleration of W on laboratory frame depends on the time laboratory frame so:
(2) L=\frac{\sqrt{(1+Wx)^2+W^2T^2}-\sqrt{1+W^2T^2}}{W}
This same equation is suitable for the measurement of any rigid or woozy rod.
We now substitute (1) in (2) and calculate the x. Obtain, omitting the details that
(3) x=\frac{\sqrt{1+W^2 L^2 sh^2Wt}-1+WL chWt}{W}
The initial time point was coordinate the front
x=L, and eventually coordinate x is growing.
Sorry for my English