March 15

This report covers the last two weeks. The week of March 8 was spring break, and little was accomplished. Likewise, little was accomplished the week of March 1 in anticipation of spring break.

In our meeting on March 2, the group decided to relieve Liusong from working on the project until after the CCE. In return, he agreed to make up the lost work after that date (March 26), and also to continue attending meetings and classes (since both the Producer and the Technical Director have read The Mythical Man-Month).

Tim and Xiao decided to set March 29 as the due date for a functional SGI LibGist. We will probably not begin working on the NT LibGist in earnest until after that date. Our research into Pentium performance measures indicates that the NT system does not provide any facilities for accessing hardware counters. Theoretically, we may be able to subvert the operating system by writing assembly code. But this isn't in the contract, and nobody on the team is excited about trying, so our specific goals (say, the entries in the user manual) for LibGist-NT are still undefined.

Xiao: I went through one assembly language book and beagn to explore the program downloaded from one Harvard paper on NT performance.

Tiger: Just wrote some small amount of codes for two classes: ProfilePane and GistView. Nothing wrote for Indicator class yet.

Rob: Thus far I've tried to finish up the classes I was assigned in our randomization for the GistView application. That went reasonably well, given that the plan was dependent on input from the rest of the group which didn't go exactly as planned. Nevertheless, we turned out a reasonable start for a design. I think my hope for a bulletproof design before we write any code was perhaps unrealistic. I've re-evaluated to a degree and am now planning a somewhat less structured approach and accepting that we're going into the coding phase with a design with some known imperfections. We met during Tuesday's class time to do a walk through of the application's execution, and while we only completed part of what the app would do, we did find a number of design flaws that I believe would not have been found by just turning everyone loose to examine their design, and that of everyone else once more.

All things considered, I classify the process as a success, if a somewhat slower success than I'd anticipated. Speaking purely for myself, the exercise reinforced for me that "buy-in" is critical in formal methods. I haven't bounced that idea off the group as a whole, but probably will in our next meeting just for my interest.

I've also started coding on the next version of the viewer app. This will be the first iteration of the production version.