monday dec13

todo

timm: 
1) remove type=X,Y
2) remove xerces-1.4
3) move graph over to the right
   3 478
   3 482
   2 43
4) rewrite intro 
   4a) get rid of quasi
   4b) partial indicator of social activity in popularity. 
       expected it to be weak. found it to be surprisingly strong

5) raj: notes on size as a conflator of defects
    - only from one data set. grrr
    - much recent evidence that static code emasures work well for inspection optimization

markus: edit section 3
raj has to write/read theories of defects
raj has to write a little something on helix
raj & markus: why, what metrics
raj: unccomment LOC vs java machine lines

================

sunday 

fig1 not clear.

important to characterize the usage.

idefect counts very low. so standard methods are fucked up

can remove LOC and transindegree

nice to throw out  LOC due to lack of definition.

out-degree

clustering co-effecient

implications. plan to write tests. where to bother 
to where to write tests

related work: small regions of most criticality

popular guys get changed the most to produce new
services. most unstable. most bugs introduced.
most bugs removed.

discussion
==========
ramble implications for component-based SE
- components are not stable

test-driven development: where to start

software testing in general

popularity extend beyond java.
c-sharp have similar semantics so may have same results

maybe early on- in a uml diagram

more case studies on larger industrial data

apply to eclipse. can see if their test suites
are focused adequately

???
LOC 3rd party stuff- ancillay classes, test classes

validity. nothing from inner clases.

todo
====
markus has to regn the data
timm has to send oo metrics paper
timm has to run 33 data sets
markus: edit section 3
raj has to write/read theories of defects
raj has to write a little something on helix
raj & markus: why, what metrics
raj: unccomment LOC vs java machine lines