In fact the code of transcriber is so nicely organized
that I've implemented already.
The idea was the following :
I need sometime to resync groups of messages that are
well in sync the one relative to the other but unfortunately there is a shift between the origin of sound and the orig of transcript.
So I wanted to push the marks to the right or to the left
introducing the notion of a 'trim segment'. the trim segment is to be shorten until a limit (epsilon) and can't be reduced more.
Some times I need to 'make room' among the segmentation
and for that the original behaviour was useful but I introduced a notion of big segment that can be trimmed
preserving the size of other segments.
Some times I don't care and the original behaviour suffice.
Now This 3 behaviours are made choosable through the context menu of the signalshape.