JFP revisions: tone down claim about subsuming KPNs

[lkuper]

Both reviewers brought this up. It's true that the claim is a bit strong!

This is mentioned in both the intro and the conclusion, if not in other places.

[rrnewton]

working on this one now, and then number 6

[rrnewton]

The objection is fair and a good observation.

Actually, just IVars get you pretty close to KPN. Cons-lists with IVar tail pointers allow multiple blocking readers, and also ensure the single writer property dynamically. (Two writers that attempt to fill the same point in the stream cause a multiple put exception.)

The missing piece is that there's no flow-control / back pressure in such a model. To do a practical implementation of KPN still needs some extra piece.