Hi Doug:

I would like you to switch your attention to the DJ paper
for which we have the input now. Our goal is
to work both towards the final reflection paper
and the planned journal submission.

Please do the following:

The information is in:
/proj/adaptive/www/papers/DJ-reflection/supplements

1. Retypeset the paper in the LNCS format. Howe much longer is it than 
   8 pages?
2. Collect performance numbers for DJ. What is a good experiment?
   Do an adaptive method with DJ and the same method with DemeterJ.
   Run the method 1000 times and compare the execution time?
3. First add all the new text that the reviewers requested. This will make
   the paper longer but we will reuse this for the journal version of the paper.

   Please focus on new material that should be added. I think the reviewers
   did not appreciate the power of asList and it should be explained better.
   This is new and has not been in earlier versions of Demeter. We had to create
   an intermediate temporary list that did not allow updates in place.

   asList allows us to define a list that cuts across a composite object.
   We can define various list views of an object by writing traversal strategies.
   An adhoc implementation of such a list view would cut across many classes
   and asList allows us to neatly encapsulate such list views.

4. Make a new copy of the paper files for the journal version of the paper.
   
5. Make a plan what to cut.

6. Implement cut and send paper in.

time line:

1. May 25.
2. May 29
3. May 31
5. June 1
6. after discussion with me on June 4.


I will be at the DARPA PI meeting next week. No meeting on Tuesday.

-- Karl