With reference to my older question Product of ramification groups: Is $G_i= H_i H'_i$?, I want to specify it in the totally ramified case:
Let $L/K$ be a finite Galois totally ramified extension of local fields with Galois group $G$. Suppose we have two linear disjoint Galois subextensions $E/K$ and $E'/K$ of $L/K$ with $EE'=L$. Let $H=G(L/E)$ and $H'=G(L/E')$. Then we get $G=H H'$.
Now consider the higher ramification groups $G_{i}= \{ \sigma \in G \mid v_L(\sigma x -x)\geq i+1\}$ where $\mathcal O_L = \mathcal O_K[x]$. $H_i$ and $H'_i$ are defined similarly.
Is it now true that $G_i= H_i H'_i$, if $L/K$ is totally ramified?
The counterexample in the referred older question uses the fact that the composite of two totally ramified extensions is not totally ramified in general.
Now this method can not be used to construct a counterexample. Has anybody an idea for a counterexample or a proof of the assertion?