Jump to: navigation, search

Difference between revisions of "Architecture Council/Meetings/September 13 2012"

(New Topics)
Line 71: Line 71:
  
 
* '''Regrets:''' Oliver Cole (standing conflict), Wenfeng Li (standing conflict)
 
* '''Regrets:''' Oliver Cole (standing conflict), Wenfeng Li (standing conflict)
 +
 +
* '''Signed-Up''': Sebastien Gandon, Wayne Beaton, Mary Ruddy, Michael Scharf, Achim Loerke, Martin O, Steffen Pingel, John Arthorne, Tom Watson, Neil Hauge, Brian Payton, Chuck Bridgham
 +
* '''Regrets''':
  
 
<!--
 
<!--
Line 88: Line 91:
  
 
=== New Topics  ===
 
=== New Topics  ===
 +
==== Platform 4.2 Performance ====
 
* Current discussions about 4.2 performance and lack of Platform committers
 
* Current discussions about 4.2 performance and lack of Platform committers
 +
** Martin: From [[Eclipse/PMC]] point of view, couple of unfortunate coincidences (1. IBM performance lab had to be closed; 2. Loss of Platform committers; 3. Juno Milestone progress looked good, so some surprise about issues being seen now)
 +
*** Good progress being made on resolving issues with CBI, Foundation Hardware for reinstating Performance tests
 +
*** What else can we do to improve the Platform staffing
 +
** [http://dev.eclipse.org/mhonarc/lists/cross-project-issues-dev/msg08123.html Cross-project Thread]; {{Bug|389215}} lower barrier, {{Bug|389175}} dual stream
 +
*** few people complaining very loud, but are we aware of a broader user community complaining?
 +
*** Quick google for "Eclipse Juno slow performance" brings up quite a bit; searching "Eclipse Juno" finds Performance related postings in Forums relatively soon
 +
*** Things do not seem overly bad but we shouldn't discount them too much
 +
** Wayne: Suggestions on the bug are good, but the feeling of a "sky is falling" kind of attitude does not really seem appropriate
 +
** John: Surprised that there was so much focus on the Performance tests ... not aware of any other project even doing performance tests, and unsure whether the tests would even have helped in the concrete cases observed
 +
*** {{bug|385272}} Things are good initially, but get really slow after 8 hours usage or after certain plugins are installed
 +
*** Perftests never found leaks or issues observed after a long time only ... perftests are valuable and important but wouldn't have found everything
 +
** Not so much loss of functionality, but harm being done to the Community by the bad buzz going on - quick communication is needed
 +
* '''RESOLUTION:''' quick communication to those involved in the buzz;
 +
** John: Communication some time next week (this week has been much data gathering)
  
<!--
 
==== Improving IP process (John A) ====
 
* Part of John's engagement as committer rep on the board
 
** {{bug|381105}} How to improve the process for contributors doing just a one-line fix
 
*** Today, 3 statements are needed on bugzilla (wrote myself, did not reference other code, have the right to contribute)
 
*** We now have bugzilla, and gerrit
 
** "signed-off-by" Field in git commit: contributor confirms to the 3 statements
 
*** Concern: How to ensure the 3 statements have actually been read and understood? Better Web Interface for contribution ?
 
*** Idea: Adding 3 statements into the commit message - having to type the text is better than just signing off on something unknown
 
  
* Example: Google Summer of Code Student
+
* Idea: Run Perftest infrastructure not only against Platform but against EPP Package ?
** Multiple / Frequent contributions
+
** Test scenarios would need to be added to ensure that additional plugins actually do have an impact
** Wayne: Why not clone the git repo, work in the clone, then contribute all in one ? (Want to test in between)
+
** Once the perftest infrastructure is on the Foundation Servers, other projects may be able to make use of it
 +
** Steffen: This would really help for Mylyn !
 +
*** John: Back-end piece for databases / perftest data correlation is some more work, not sure yet who can spend time there
 +
* '''RESOLUTION:''' improve automation on Eclipse.org hardware (CBI, Hudson, perftests, Gerrit, ...)
  
==== Logging and third party libraries (John A) ====
 
* 2 areas, (a) error loging, (b) tracing.
 
** Some 3rd parties use slf4j, logback, ...
 
** Igor: slf4j (or its configuration) should be part of the Platform
 
*** Today, projects (like m2e) have to ship a logging configuration, but that may conflict with others trying the same
 
*** logback seems to be the most common configuration
 
*** There should be a way allowing projects contribute to the configuration
 
** {{bug|358968}} - suggestion to link other bugs from that bug
 
** Logback implementation not really designed for OSGi ?
 
  
* Martin: Related discussion - Eclipse Standard tracing API's (brought up by Tim a while back)
+
==== Upstaffing Platform Committers ====
** John: Eclipse Platform added dynamic tracing API like 2 years ago (dynamically turn on/off tracing, new Prefpage for users as of Juno)
+
** Igor: slf4j is not a logging framework - just a thin adaption layer; logback is one particular back-end
+
** What API's being used is not so much of an issue - the question is how to properly wire up back-ends
+
  
 +
* Martin: Ideally, companies / organizations need to invest, allocating full-time committers
 +
* Martin: How to make best use of those people who are investing "some" time (eg for a patch) but not full time ?
 +
** Wayne: Could Mentors / AC Members / Community in General take a bigger role in the Review process ?
 +
* Michael S: There need to be some people who have the deep understanding and vision to maintain the Platform architecture and vision - we need more of those.
 +
** John: The most helpful thing right now is LTS - Companies contributing committers for fixes across everything ("shallow") with focus on stability
 +
** Lowering the Barrier: The biggest part is how much effort the contributor puts in (range from just identifying the bug - giving bug+fix+patch+tests)
 +
** Martin: So there is space for a role of a "Bug Mentor" that helps contributors go the extra step also submitting explanation + tests to get a "greatbug"
 +
*** Could this be automated somehow ?
 +
** Currently there is just 1 giant eight-hour test suite, this should be broken up somehow
 +
* How is the "tragedy of the commons" situation with other Communities? - Eg Apache seems similar to Eclipse, staffing varies greatly by project
 +
 +
* '''Next Steps?''' eg re Mentoring bugs
 +
** How to identify those bugs that are interesting to look at ?
 +
** John: Bugzilla queries like "attachment has patch" worked in the past, but identifying comments with a git URL is harder
 +
*** '''AI Wayne''' check whether could hack up something
 +
*** Martin: Maybe add a "comment is patch" flag on the data entry side ? Eg Review flag ?
 +
 +
 +
==== Announcements ====
 +
 +
* Wayne: Hitting the T-3 Months mark of CVS retirement.
 +
** Please inform mentored projects and make them plan for git migration. Website migration is possible now {{bug|324116}}
 +
** Git Migration Status of Projects:
 +
 +
 +
 +
<!--
 
=== General Topics  ===
 
=== General Topics  ===
  

Revision as of 11:01, 13 September 2012

Meeting Title: Architecture Council Monthly Meeting
Date & Time: Thursday September 13, 2012 at 1500 UTC / 0800 SFO / 1100 Ottawa / 1600 London / 1700 Berlin attention DST change
Html.gifHTML | Ical.gifiCal
Dial-in: Let's use the Foundation's Asterisk setup for this call:
  • Ottawa (local call in Ottawa) 1-613-454-1403
  • North America (toll free) 1-877-369-7806
  • Germany (local call anywhere in Germany) 49-692-2224-6059
  • France (local call anywhere in France) 33-17-070-8535
  • UK (toll free) 0800-033-7806

Participant conference extension: 701 then enter pin: 51968

  • SIP clients can call 701@asterisk.eclipse.org, then enter pin 51968.

Attendees

All AC Members are invited.

  • PMC Reps please confirm attendance or list your delegate below. Every PMC is required to name a primary and backup delegate, and to ensure that one delegate attends the meeting.
BIRT: Wenfeng Li Wenbin He
DTP: Brian Payton Linda Chan
Eclipse: Mike Wilson John Arthorne
Boris Bokowski
Modeling: Ed Merks Cédric Brun
Sven Efftinge
Mylyn: Steffen Pingel Mik Kersten
RT: Christian Campo Tom Watson
SOA: Adrian Mos Sebastien Gandon
Technology: Gunnar Wagenknecht Wayne Beaton
Tools: Doug Schaefer
WTP: Chuck Bridgham Dave Carver
Neil Hauge
  • Regrets: Oliver Cole (standing conflict), Wenfeng Li (standing conflict)
  • Signed-Up: Sebastien Gandon, Wayne Beaton, Mary Ruddy, Michael Scharf, Achim Loerke, Martin O, Steffen Pingel, John Arthorne, Tom Watson, Neil Hauge, Brian Payton, Chuck Bridgham
  • Regrets:


Agenda / Notes

  • Feel free to edit, but not during the call!

Review of Last Meeting

New Topics

Platform 4.2 Performance

  • Current discussions about 4.2 performance and lack of Platform committers
    • Martin: From Eclipse/PMC point of view, couple of unfortunate coincidences (1. IBM performance lab had to be closed; 2. Loss of Platform committers; 3. Juno Milestone progress looked good, so some surprise about issues being seen now)
      • Good progress being made on resolving issues with CBI, Foundation Hardware for reinstating Performance tests
      • What else can we do to improve the Platform staffing
    • Cross-project Thread; bug 389215 lower barrier, bug 389175 dual stream
      • few people complaining very loud, but are we aware of a broader user community complaining?
      • Quick google for "Eclipse Juno slow performance" brings up quite a bit; searching "Eclipse Juno" finds Performance related postings in Forums relatively soon
      • Things do not seem overly bad but we shouldn't discount them too much
    • Wayne: Suggestions on the bug are good, but the feeling of a "sky is falling" kind of attitude does not really seem appropriate
    • John: Surprised that there was so much focus on the Performance tests ... not aware of any other project even doing performance tests, and unsure whether the tests would even have helped in the concrete cases observed
      • bug 385272 Things are good initially, but get really slow after 8 hours usage or after certain plugins are installed
      • Perftests never found leaks or issues observed after a long time only ... perftests are valuable and important but wouldn't have found everything
    • Not so much loss of functionality, but harm being done to the Community by the bad buzz going on - quick communication is needed
  • RESOLUTION: quick communication to those involved in the buzz;
    • John: Communication some time next week (this week has been much data gathering)


  • Idea: Run Perftest infrastructure not only against Platform but against EPP Package ?
    • Test scenarios would need to be added to ensure that additional plugins actually do have an impact
    • Once the perftest infrastructure is on the Foundation Servers, other projects may be able to make use of it
    • Steffen: This would really help for Mylyn !
      • John: Back-end piece for databases / perftest data correlation is some more work, not sure yet who can spend time there
  • RESOLUTION: improve automation on Eclipse.org hardware (CBI, Hudson, perftests, Gerrit, ...)


Upstaffing Platform Committers

  • Martin: Ideally, companies / organizations need to invest, allocating full-time committers
  • Martin: How to make best use of those people who are investing "some" time (eg for a patch) but not full time ?
    • Wayne: Could Mentors / AC Members / Community in General take a bigger role in the Review process ?
  • Michael S: There need to be some people who have the deep understanding and vision to maintain the Platform architecture and vision - we need more of those.
    • John: The most helpful thing right now is LTS - Companies contributing committers for fixes across everything ("shallow") with focus on stability
    • Lowering the Barrier: The biggest part is how much effort the contributor puts in (range from just identifying the bug - giving bug+fix+patch+tests)
    • Martin: So there is space for a role of a "Bug Mentor" that helps contributors go the extra step also submitting explanation + tests to get a "greatbug"
      • Could this be automated somehow ?
    • Currently there is just 1 giant eight-hour test suite, this should be broken up somehow
  • How is the "tragedy of the commons" situation with other Communities? - Eg Apache seems similar to Eclipse, staffing varies greatly by project
  • Next Steps? eg re Mentoring bugs
    • How to identify those bugs that are interesting to look at ?
    • John: Bugzilla queries like "attachment has patch" worked in the past, but identifying comments with a git URL is harder
      • AI Wayne check whether could hack up something
      • Martin: Maybe add a "comment is patch" flag on the data entry side ? Eg Review flag ?


Announcements

  • Wayne: Hitting the T-3 Months mark of CVS retirement.
    • Please inform mentored projects and make them plan for git migration. Website migration is possible now bug 324116
    • Git Migration Status of Projects:



Action Items

  • Cleaned up old action items, see Architecture Council/Meetings/February 10 2011 for old stuff
  • (old) Martin to add Eclipsecon meeting notes onto the wiki
  • (old) Tim write up an initial wiki page with information for people to standardize on the tracing API
  • (old) Martin revise the AC Wiki to make it easier to find the New Member Process. More links on homepage. More usage of categories.
  • (old) Martin bug 315210 Make the AC mailing list open / moderated
  • Martin contact / propose potential new AC members

Next Meeting