| Meeting Title:
|| Planning Council Conference Call
| Date & Time:
|| Wednesday, March 07, 2012, at 1200 Eastern
|| For the call-in numbers, see the "Project Review" number on Foundation Portal page.
Members and Attendees
PMC (and Strategic) Reps
| Chris Aniszczyk
|| Technology (PMC)
| John Arthorne
|| Eclipse (PMC)
| Mik Kersten
|| Mylyn (ALM) PMC
| Brian Payton
|| Datatools (PMC)
| Doug Schaefer
|| Tools (PMC)
| Adrian Mos
|| SOA (PMC)
| Ed Merks
|| Modeling (PMC)
| Jesse McConnell
|| Rt (PMC)
| David Williams
|| WTP (PMC) (appointed Chair)
| Gary Xue
|| Birt (PMC)
| Wayne Beaton
|| Eclipse Foundation (appointed)
| Cedric Brun
|| OBEO (Strategic Developer)
| Stefan Daume
|| Cloudsmith Inc.(Strategic Developer)
| Neil Hauge
|| Oracle (Strategic Developer)
| Kaloyan Raev
|| SAP AG (Strategic Developer)
| Igor Fedorenko
|| Sonatype (Strategic Developer)
| Markus Knauer
|| Innoopract (Strategic Developer)
| Christian Kurzke
|| Motorola (Strategic Developer)
| Achim Loerke
|| BREDEX (Strategic Developer)
| (PMC rep)
|| Actuate (Strategic Developer)
| (PMC rep)
|| IBM (Strategic Developer)
| [no name]
|| TPTP (PMC)
| [no name]
|| CA Inc. (Strategic Consumer)
Note: "Inactive" refers to Strategic Members or PMCs we have not heard from for a while, and have been unable to convince to participate. Those members can become active again at any time. Contact David Williams if questions.
Note: feel free to correct any errors/omissions in above attendance record.
Y = Yes, attended
N = No, did not
R = regrets sent ahead of time
D = delegated
X = not expected
- Welcome Igor Fedorenko as new Sonatype Strategic Developer Representative.
- Any others?
- None discussed (but, send note to mailing list, etc., if anything occurs to you later.
- Issue to discuss and decide if we need a plan of action: p2 content metadata at SR2 is 3 times what it is at SR0.
- Should our common "release repo" contain only the latest code? And "move" older stuff to different site? The different site would be named something like .../releases/juno/complete or or something, and be simply a different composite (no duplication of actual artifacts). This site would not be "built in" to any update repo lists, but could be used by builds or others that needed "the old stuff". I think if someone updated, and then wanted to revert or rollback the change, they might also have to manually add the ".../releases/juno/complete" URL to their list of sites (assuming p2 GC had cleaned off the old stuff. Please read the msg chain on p2-dev list. There is a trade off of function and performance and want to be sure everyone is aware of it and if anyone has any opinions on if we currently have the right choice.
- Another solution might be to aggregate new stuff, with old, which would have added advantage of making sure all was compatible ... but, a) not sure it would be much smaller and b) requires all projects to do an "expert job" of versioning.
- No strong view to change, so normally status quo would rule. But, we did discuss it is a little hard for us to know magnitude of problem, and if it'd be worth trading off the current function, so, a bug might be worth opening, and see if some "world wide" data was worth collecting. To us, at meeting, there's a big difference if its 90 seconds for most people and the would improve to 30 seconds. Versus 15 minutes for most of world, thus reducing to 5 minutes. Concern was also expressed that it may be an inherent problem/limitation with p2 and this "fix" would be sort of "temporary" ... that if future repos grew to three times current size, we'd be in the same "poor performance" category, unless p2 was improved or some other solution found.
- No bug yet, but I've gotten some email one might be open to "remove categories" from Indigo SR0 and SR1 repositories, to "complete" the fix for the problem caused by linuxtools changing feature IDs in SR2 (See bug 371302). As is, 6 features show up in "Linux tools" category, but 3 are for SR1 and 3 are for SR2 and they can not be all installed (all at once, that is, as most users would "pick them all" if they wanted Linux tools). While not exactly a blocking bug ... is it something that will last for years to come ... not to mention, I wonder if we should always only have one categorization for common repo? (discussed some in bug 314165.
- Any issues/thoughts on this? Allowable? Desirable? Not?
- Will leave as "for awareness" until/unless bug open. John did say, "he didn't know about magnitude of the bug, but that he thought is could be done safely, since its just used in UI" ... not a fundamental IU used in builds or installs, or anything. I guess anyone who had made "internal mirrors" might need to redo?
Ready for M6?
- Sounds like fair progress towards moving to "non greedy" publishers.
Issues or Exceptions
Any issues? Everyone in? Any exceptions known?
- Exceptions for projects not in M4, that still will to join Juno:
- Virgo approved during 1/05 meeting (from rt PMC list, will be in M6)
- BPEL approved on mailing list (as late for M4, but in M5)
- Code Recommenders approved on mailing list (as late for M4, but in M5).
- Koneki project approved on mailing list (as late for M5, but joining in M6).
Anyone "dropping out" that should be removed from aggregation build?
- removed following b3aggrcon files, for M6:
What to do about Papyrus (and non released XWT dependency)
- Both in general (bug 370974), and specific for this case.
- In general, can I say the Planning Council agrees with my summary conclusion in the bug 370974#c12?
- Consensus was "yes"
- Specifically, does the Planning Council agree this means Papyrus can not be in Sim Release? In fact, could not release at all, until this XWT issue is resolved.
- They perhaps could use an "old" version of XWT? But, my guess is that very old release (which happened sort of erroneously) does not have about.html files, etc.
- XWT could graduate/move to its own project and release from there ... but not much time left, and seems unlikely?
- They could also release/join during SR1, etc., if things work out by then.
- Keep in mind, one bad aspect of this is that Papyrus and some "unreleased" XWT bundles were in Indigo. We should have "caught the error" back then.
- Agreement they can't release (much less be apart of train) with unreleased dependencies. Wayne said there could maybe be exceptions, but it would have to be a very special case, under just the right conditions. Also, it was said, that given that "it was done wrong last year" speaks to the magnitude of their problem ... that is, that it is still not fixed/released by now, is even more reason to not allow it to happen again.
- ACTION: dw to send note to papyrus and modeling pmc saying "no", and remove them from aggregation build, unless they know something we don't (e.g. I'm pretty sure XWT is central to their function, but for all I know maybe they could just leave it out, and haven't said so yet.
- Does anyone suspect any other, similar cases?
- None mentioned.
- anything to look at? In particular, plans specifying "planned support for 3.8 workbench"?
- no consolidated report yet
- The proposal (to summarize, assuming no communication problems) was for the "bits" in Eclipse Classic, and the Juno repository to be provided by the Eclipse releng team.
- But, that the bits used in EPP (for the platform/SDK bundles) come from the CBI builds, and the rest come from common repository.
- Issue one: this means there would be two versions of platform bundles "in the wild" for the Juno release.
- Either with same version and qualifiers, but (probably slightly different content), or different qualifiers but maybe same content ... either case leading to chaos (IMHO).
- I think "updates" and "what would adopter's customer's end up with" are huge problems for Eclipse, in these cases.
- Issue two, while is is true, Eclipse is open source, and "people can do with it what they want", I think it is entirely up to the EPP project what they do and how they do it (not "CBI") and such a large technical/process change should get plenty of discussion with EPP Project (and, probably their PMC ... is this something they think is a good idea?)
- Issue three, it seems all this is happening "last minute" instead of the proven "Eclipse way" of producing milestones, making steady, incremental progress towards a release.
- Issue four, would projects still want to be in EPP, if they knew bits would be different than if they "built their own"?
- Issue five, I think the Planning Council is in charge of "the release plan" (since, it effects all top-level projects and PMCs and Strategic Members) so I think we are obligated to make a statement on this issue.
- If not obvious, I think we all want to be supportive of CBI and the LTS efforts, but ... even if bundles are planned/considered to be identical, then why have two versions of them? Part of the answer is to make very rapid progress in CBI and LTS. Again ... I think we all support that rapid progress, but I'd prefer a plan that had same bits, all around. Perhaps SR1?
- Any thoughts? Suggestions for what our Planning Council statement should be? Can we make a concrete statement or recommendation?
- Thanks to Andrew Ross for going the call as a"special guest" to answer questions and hear concerns and learn more about bundles and our normal way of ordinating a release.
- Agreement that having "two versions" of bundles for one release is not tenable (for maintenance, predictable builds, updates, etc.) and would tarnish Eclipse's reputation for producing reliable, predictable releases. In some ways, it was speculated, that maybe the proposal made on cbi-dev list was just a case of partial mis-communication or terminology issues of not understanding the nuances of repositories, installs, updates, adopter scenarios, as well as the Eclipse dev. process and the project decision processes.
- Markus did say he might be willing to try using CBI output in some "test builds" for evaluation, if anyone thought that would be useful, but would not want to "release" those, if the platform released something different in the common repository. The EPP packages all come from the "common repository" which he does not want to change -- that is a key part of the "coordinated" and "update" story at Eclipse. And, it is up to each project to decide what goes into that common repository (that is, they decide what they contribute), since it is their responsibility. And, the Platform team has decided that they will contribute the bits from the PDE build for SR0 in June ... but they are willing to consider changing build systems after June, if CBI proves ready by then, or even after SR1 (for SR2 release) if CBI not ready by September or not enough time to evaluate it by September.
- The Planning Council agreed it would be a better plan to work "bottom up", to first prove the CBI can build the same thing as the platform build, and to then move the platform to a CBI build, and then end up with one set of bundles per release. It was felt that this might be feasible by SR1. Even that would be aggressive, and have some risk of "making it on time" (without sacrificing quality of the builds produced) but would feasibly allow time some iteration and investigation and proof that it was adequate if not identical, before committing bundles "out in to the wild". Perhaps for Juno, "LTS" could "start with" SR2 ... though long term, Andrew told us, LTS members do want to be able to have their own builds/fixes/streams even immediately after SR0, if they so desired.
- So, our recommendation is for CBI and LTS interests to come up with a plan to get the platform to move to CBI build by SR1 or SR2, and not plan to release two sets of bundles, as well as work with the EPP project to make sure the project (and the package maintainers and committers) agree with any proposals made to change the way things are done.
- Planning Council members (or, members of their teams) can support the CBI and LTS efforts by at least opening bugs/feature requests on what's needed for CBI to "work for them" (in cbi bugzilla product), be explicit about acceptance criteria, as well as contribute man power to try the builds, test the results, etc., if not actually provide tools for CBI (the Common Build Infrastructure).
- ACTION: dw will post "recommendation" to CBI dev list for proper visibility.
- Project Priorities: Please review and be prepared to discuss this proposed "policy document" about project priorities.
- One issue: should we mention LTS? Technically ... it is not in our mission or scope.
- Are we in agreement these can be published a "priorities from Planning Council's point of view" to begin "socializing" the idea?
- Will Java pack200 issue bug 361628 need action? Is a fix possible? It will not literally be a problem if everyone published both jar.pack.gz files and jars, but would be inefficient (ending up with "failures" with pack.gz files, and then downloading jars if using Java 7). Keep in mind, we have decided in the past that we should always publish both *.jar and *.jar.pack.gz files ... so, no change in policy for Sim. Release.
- EclipseCon face-to-face meeting: Sunday, March 24, 2 - 4 local time (Eastern). Joint meeting with Arch. Council 4 - 5.
- Agenda will be developed soon, but good time to discuss Kepler and other "big picture" items.
- April 4, 2012 (our regular "first Wednesday" meeting, at Noon Eastern).
- Juno Wiki page
- Planning Council/Indigo retrospective
- Planning Council Members
- Simultaneous Release Roles and Simultaneous Release Roles/EMO