1: \begin{proof}

2: %\mv{

3: %Would be nice to have a-priori proof that $\frac{r}{2}$ nodes moves to $b$ and their weights vanish, but I'm afraid it is too tough.

4: %I got stuck here, did not realize we have mixed continuities. If they are uniform, there must be at least $\frac{r}{2}$

5: %nodes inside $[\XXt_n[K-r],b]$ (this is Micchelli's theorem about the nodes' range) so the first part is easy.

6: %Well, proof, everything is under a strong assumption that the continuation traces optimal rules, this actually must be also

7: %a part of the assumptions.

8: %

9: %The second part (weights) is to me non-trivial, the last $r$ equations of the system become $0=0$ in the limit,

10: %so one cannot conclude that $\omega_{m-\frac{r}{2}+i}$, $i=1,\dots,\frac{r}{2}$ goes to zero.

11: %An observation I have is that if, e.g. (for single interior knot $x_i$ moving towards $b$, let $h=b-x_i$) the basis function

12: %behaves like $x^2$ in the neighborhood of $x_i$ and the evaluation is therefore $\mathcal O(h^2)$, but the integral changes as $\mathcal O(h^3)$

13: %and therefore $\omega$ should behave as $\mathcal O(h)$, i.e., go to zero. But I am far away from a rigorous proof.

14: %But again, this all is under two assumptions: we trace optimal rule and the multiplicities are uniform,

15: %(which we don't have for the knot transformation described in Section~\ref{sec:splines}).

16: %

17: %To me it is quite important to know in advance that $\omega_i\rightarrow 0$, because one can disregard the last $r$ equations

18: %of the system (or consider the full $2m \times 2m$ system and set them $\tau_i = b$, $\omega_i =0$, which is done in the implementation,

19: %otherwise Maple starts screaming that the system is not well-defined). }

20: %\end{proof}