Thesis Modifications
Architecture of Fine-Grained Data Flow Network Programming environments

Michael Yuh-Jye Chang Thesis July 23

There was general agreement that the technical quality of the work was exceptional
but that the presentation could be improved.
As advisor, I consider it unfortunate that there are no published papers to record this
work in computer science literature. I would personally recommend strategy
a) Write Paper
b) Put material from Paper into thesis
This makes implementation sense because the type of material needed to
improve thesis is precisely type of material that is needed in thesis

There are some marked up thesis copies which I will send. 
My request is simple
There are many sections called "conclusions".They are not conclusions but
rather summaries. Conclusions should be precise clear lessons. There are
a few of these but not many.

The reader should come away from thesis with a clear understanding of
a) What Neat Tools is meant to do
b) What design decisions were made in Neat Tools
c) What good bad and/or pragmatic decisions were made in implementation
d) Relationship of Neat Tools to other systems
e) If you were starting today (or two years from now) what should one do
	to build a system that has certain design goals which could either be
	same as those for current system or modified to include goals used in AVS,
	LabView or totally new requirements etc.

Currently I believe there is a good description of say LabView but not 
of what I asked above. These observations a) through d) should either be
in chapter 5 or in subsidiary sections like 2.2.4 etc. (anything called conclusions)

Here are some Specific Comments I noted. There are more in marked up thesis copies

1) Could address several issues in paper(s):
   Java Features from C++
   Fine-Grain Data-flow
   Medical Use
   
2) pp.7-8: Update results Java2/Hotspot
   C++ may be better for YJC but it is claimed by Sun to be 50% worse 
   for average weaker programmer
   Java frameworks (i.e. libraries like JDBC) are important
  
3) What about buffers/reliability etc.These seem to favor Javabean event model
   over faster synchronous messaging. More generally need a discussion
   of handling of race conditions, real-time contraints (data not ready,
   buffers overflow etc.)

4) Discussion in thesis and exam of "Is Neat-Tools a CPN (Petri Net)"
   was not very precise
   YJC says "Yes!! It is much faster"
   I did not see this demonstrated
   Any CPN is a Neat tool seemed possible
   Any Neat Tool Application is a CPN -- less plausible
   
5) Pages 41 and 76 need more explanation
   as to precise relationship of UML and Neat Tools