Map folds in containers: why no worker/wrapper split?

[wren ng thornton]

Claus Reinke wrote:
> Lots of performance problems with containers are about
> pragmatics, not representation, eg, one knows that Maps are
> strict in keys but available operations happen to be defined
> in such a way that one cannot easily ensure other desirable
> strictness properties.

Not to put any extra burden on the messenger, but do you have a summary 
of the particular strictnesses that are often desired, or the common 
problems encountered in practice?

In particular I'm thinking in the context of the bytestring-trie package 
and how it can be improved. The initial API sought to mirror the 
containers version of Map/IntMap, because they seemed rather stable at 
the time. However, there've been a number of changes and proposals 
lately, so I want to be sure I don't miss anything.

-- 
Live well,
~wren