IMPORTANT
---------

Q: Does it matter if you do CbX vs CbY?  
A: Only in the matters of fine print (the complexity results will
certainly remain unchanged).

THIS SHOULD DEFINITELY BE DISCUSSED AS CBV vs CBN has been dealt with
in several papers.

Q: What happens if you closure convert the kCFA EXPTIME construction?
(Olivier's ICFP'07 question)
A: Dunno, but the completeness results tells us:
   Either you can't answer flow questions about the original program or
   you still have to do an exponential amount of work.

THIS SHOULD DEFINITELY BE DISCUSSED AS flow analysis and closure
conversion have been dealt with together in several papers.

[Also: how are you going to represent the "records" the will be used
to represent procedure values in the closure converted program --- you
only have \lambda, so you'll Church encode a tuple.  In other words
you'll use closures to represent records representing closures.
You'll still have closures in the analysis.

Alternatively: consider closure conversion into a first order language
with records and a traditional data flow analysis.]

LESS SO
-------

Change theorems to use small caps, no bold, no italics (see CMS
13.44).

Remove junk from bibliography, ie. DOI, ISBN, etc. ?? (This can be
done by creating a custom bibliography style using latex
makebst. Remember to use author-year citations and natbib.)

Work in Levy citation from Curry festschrift.  Should be in discussion
of optimal evaluation.

========


Watch out for: EXACT, EXACTLY
(over-used, over-loaded)


What is an {\em exact} flow analysis?

Be consistent with "the flow problem", "the control flow problem",
etc.

Q: What exactly is an "Axiom wire" correspond to in a program?
A: (Don't say variable occurrence).

Connection with Danvy work on CPS, defunctionalization.  Can this be
applied to an {\em abstract} interpreter?

Give a one sentence eplaination of Theorem 1 in 0CFA; what do the
preconditions mean?  It says that \ell can't already be in use in the
cache or appear in the current term or environment.

Look into flow analysis based on intersection typing.  Apply the
Neergaard work to this.  What can be said about the expressiveness of
these analyses?

Cite and discuss: A New Approach to Control Flow Analysis, Pasquale
Malacaria, Chris Hankin.

Also Malacaria and Hankin point out that extending Heintze and
McAllester's analysis to a language with constants and conditionals
results in a cubic analysis.

Be consistent with italics and quotation environment.

Cite and discuss Mackie static analysis of interaction nets.

Citation for MCD and exponential blow up of MCD.  Barendregt?  Levy?

PL Theory notes on MCD.  MCD description of Neergaard Mairson ICFP 04.

Use of MCD in strong normalization proofs.

Might's dissertation should have further discussion.

In 4.4.2, Should first describe the GoI algorithm for non-atomic
axioms, prove it to be in P, then do the stackless algorithm.

Insert nCFA stuff from ICFP'07?

Consistent with sec- section- label prefix.