Coordination and frames

In investigating solutions to handle the nested form issue we discussed 
at the F2F I have been looking into using frames instead of form 
rewriting as a potential solution.  This has lead me down the track of 
trying to figure out how our coordination could work when a page is 
rendered as a set of frames.  The basic problem relates to our 
assumption that coordination is scoped for a set of portlets [a page] 
yet few portlets assume they are rendered in frames hence don't 
manage/set the link/form targets in their markup yielding the default of 
"self".  I.e. the typical portlet's link and form refs will result in 
the browser filling this portlet's frame with whatever content is 
returned; for correctness this must be a link/form reference that only 
renders the referenced portlet. 

So how does coordination work in this environment?  Public parameters 
and events are designed to interact with other portlets on the page 
resulting in these other portlets also needing to render.  But as said 
above, we typically in a conetxt to rerender/receive these additional 
renders.  Ideas? 
    -Mike-