Discussion Notes SQ 2009 Week 5

One of the interesting things of the paper is the diagram adapted from the paper of YE and Fischer on page 4. We realize that there is an important manner in which developers carry out reuse. Often there is reuse by anticipation, where you think that a particular thing should be there. It is directly related to the level of ignorance. SO, if it is a component library, it is too big to know what might exist there. Reuse by anticipation can also be termed as reuse by guessing. The very idea of anticipation, but not knowing exactly whether the thing you are looking for exists, is interesting..

We discussed how we realized that this paper was a typical AI paper where they promised a lot, and delivered very little. It is similar to when people try to use 3D environments without actually knowing how to use it. They just take blocks in 2D and change it to buildings in 3D.

We say that something “really works” if someone who is not the creator/implementer of the tool picks up the tool and is able to use it.

There was a discussion about the diagram on pg 8, We tried understanding code repository and component library. The Component Library contains what is being reused for e.g Swing, AWT but we think that even component lib should have its own code repository.

The code repository on the other hand contain actual examples like the code repository could be source forge.

Between the Code repository and Component Library, item get mapped to actions.

In Fig 5 on pg 9, we see that in the signatures, they are the raw signatures which might be incomprehensive to some people. But that is a minor problem.

For the model based algorithm, they have not given which metrics should we be clustering on.

Figure 8 on page 14 is an interesting figure from which we see the order in which the methods were called. It involves matching similar sequences and determining what is the next method to be called.

In the evaluation part on page 18, there is no chart for time. It is hard to see why they have taken F1 to be 2PR/P+R. So basically it is not clear how exactly is it good to summarize P and R?

Also, in the results summary they haven’t compared base line to anything.

The paper is a nice follow on to code broker paper. Its introduction is nice and the paper gives a lot of encouragement to cut and paste programming.