1: \begin{proof}

2: Instead of a rigorous proof we provide only a sketch, the easy details are left

3: to the reader. In both algorithms we have $n$ intervals, thus $n$ steps. In each

4: step we define a bounded number of new active intervals, thus altogether we have

5: $cn$ regular and active intervals. We always maintain the (well-defined)

6: left-to-right order of the active intervals. Also we maintain an order of the

7: (active and regular) intervals such that an interval's color depends only on the

8: color of one or two intervals' that are later in this order. This order can be

9: easily maintained as in each step the new interval and the new active intervals

10: come at the end of the order. We also save for each interval the one or two

11: intervals which it depends on. This can be imagined as the intervals represented by vertices on the

12: horizontal line arranged according to this order and an acyclic directed graph on them representing the dependency relations, thus each edge goes

13: backwards and each vertex has indegree at most two (at most one in the first

14: algorithm, i.e. the graph is a directed forest in that case). In each step we

15: have to update the order of active intervals and the acyclic graph of all the

16: intervals, this can be done in $c \log n$ time plus the time needed for the

17: deletion of intervals from the order. Although the latter can be linear in a

18: step, yet altogether during the whole process it remains $cn$, which is still

19: ok. At the end we just color the vertices one by one from right to left

20: following the rules, which again takes only $cn$ time. Altogether this is

21: $cn\log n$ time.

22: \end{proof}