> I would love to be able to see the full tree of history and  
accomplish what you
> have described. Actually, I would like to be able to use such a  
capability to
> perform diffs and merges of arbitrary nodes of multiple branches and 
> proceed
> with an independent branch (I'm not sure what to call it). Basically, I 
> would
> like to pick up a bug fix from one branch, a feature from another,  
and proceed
> along a separate thread of changes. I have often had to do this  
kind of thing
> manually more than once and it tends to be a pain.

I think this boils down to a simple sort of module algebra:  'Take  
module A, transform it by taking the following diffs from module B  
and, together with module C use that.'

Right now, these are steps to be taken in a linear fashion, lost as  
steps and only visible in the changes they make in the image.  Being  
able to retract changes and go a different way is a start, but I  
think it would be even better if these steps could be sentences we  
can write down, save away, review and modify at a later date.

With this view, it also becomes apparent that change-management and  
modules are actually orthogonal issues:  change management can be  
applied to any type of object, modules just provide an aggregation  
mechanism in order to treat related objects as a single entity.

(Change management and granularity do interact quite ferociously,  
because it is often not clear at which level of granularity changes  
should be recorded in composites)

Marcel