so in trying to understand the idea of the "cartan-rolling distribution" coming from a "cartan geometry", we should in particular consider the ("constant curvature" ?) case of a homogenous cartan geometry... perhaps exploiting special algebraic techniques that become available in this case... ??and perhaps becoming particularly relevant in connection with invariant distributions on partial flag varieties??

to what extent do _all_ distributions arise as "cartan-rolling distributions"? and to what extent can the original cartan geometry be recovered from the distribution? i guess that that those are roughly surjectiveness and injectiveness questions, respectively... ??is there maybe a third question in that family somewhere?

to what extent do continuous parameters occur in the classification of graded nilpotent lie algebras?? (or something...) there's stuff here that confuses me... concerning the radius-ratio parameter in the rolling ball geometry, and so forth....

perhaps derek would prefer to call a "cartan-rolling distribution" a "generalized hamster-ball rolling distribution" or something...