Sat Oct 31 17:05:31 EDT 2015

lin types

What are lin wrappers for?
I no longer understand the point of this.

It seems that moving between graph and lin memory is very straightforward:
- convert cell type
- adjust ref counts?

Note that the graph memory only needs one RC increment.  Decrement
happens on GC.

Tried to remove, but there are all kinds of assumptions.

Needs a lot of attention to understand the basic point, and find a new
invariant.  With RC counts inside the types it shouldn't be so hard to
copy between GRAPH and LIN memory.