[eclipse-dev] Planning Meeting Notes Nov. 27, 2002 (abbreviated :-( )
The technology I was using to record the planning meeting (silently) failed
after the first half hour. As such, these notes cut off before the end of
the meeting. Sorry all. I'll have to go back to doing it the old fashioned
way (/em sighs). If there's something missing which anybody who was there
wants mentioned, please send a note to eclipse-dev.
*** General ***
- stop ship PR against 2.0.2 last week (install/update) fixed on weekend
- new book on Eclipse being reviewed by several committers
*** Good stuff last week ***
- Hotcode replace now works on the Mac, running on the latest 1.4.1 Java
- almost all milestone plans are posted
*** Focus: Performance ***
- almost all teams have submitted performance problems (93 PRs +/-) to look
- encouraged to look at cross-component problems more (some teams have
started doing this)
- do not currently have end-to-end scenarios to measure
- it would be good to be able to show that the system really is faster
for the standard
Eclipse usage patterns (including scalability)
- note: teams really do need to do their own benchmarks (should be run all
- be sure to benchmark on all platforms, both because performance
significantly and because user expectations vary as well.
- would like a reference large workspace to test scalability. Currently,
just load all
of Eclipse from source.
- what are our expectations for memory use w.r.t. scalability?
- web site for performance focus work today or tomorrow
- performance wins already:
- reduced number of label updates in tree viewers by 2x.
- fixed a Navigator bug which caused double refresh
- fixed memory leak in java editor
- can we further trim the amount of state we remember (eg. expansion state
in trees) to reduce the work at startup?
- saw an instance of tomcat threads being started at startup. should only
be on help invocation. Help team to investigate.
- spy tools are good at discovering why plugins have been activated. It
appears that several of the plugins which
are activated at startup do not need to be. All teams to investigate.
*** Outreach ***
- 2 talks at IBM developer conferences in Germany (400 people and 150
- Eclipse welcomed.
*** Status ***
- outline in content assist (control-o)
- experiments with encoding work: does not seem to be any "small" version
of the problem
(i.e. any solution has pervasive impact)
- auto-edit strategy (generalization of auto-indent work)
.... (The rest was lost. Again I apologize).