OpenJDK Community Census

mark.reinhold at oracle.com mark.reinhold at oracle.com
Mon Sep 12 14:51:31 PDT 2011


2011/9/8 17:18 -0700, ahughes at redhat.com:
>     ... I was pleased to see acknowledgement of my work by my new positions,
> but I am confused by a number of things:

Thanks for your questions.  Answers inline below.

> Groups
> ======
> 
> I am a little confused as to how groups fit in or even as to what
> their purpose is.  The bylaws defines a group as 'a collection of
> Participants who engage in open conversation about a common
> interest.'.  It's not clear to me what the point is of defining that
> collection and having membership of it, given such conversation
> regularly takes place on the mailing lists between both people who are
> members of these groups and those who aren't.

Groups are meant to capture the slowly-evolving social structures of the
Community.  As such they're generally longer-lived than Projects, and
they're also a bit harder to create and dissolve.  A Group can have its
own web content, independent of any particular Project, which is writable
by all of the Group's Members.  (So far that service hasn't been used
very much, but I expect it to become more widely used as the overall
infrastructure improves.)

Membership in a Group is a stepping stone to becoming an OpenJDK Member;
being a Committer in at least one Project is the other.  There are two
paths to OpenJDK Membership so that people who make contributions in
forms other than code can eventually become OpenJDK Members and thence be
able to vote on new Project proposals, in GB elections, and so forth.

> It's not clear how those currently not in a group become members or create
> new groups.

This is spelled out in the Bylaws [1].  Prior to the Bylaws we had
informal interim guidelines for creating new Groups and for nominating
new Group Members [2].  They were not often used.

>              Indeed, I'm not aware of any group members who aren't either
> current Oracle employees or were Sun employees at some point.  Are there any?

There are a few.  Doug Lea, e.g., is a Member of the Core Libraries
Group.

> I have reviewer status on four projects and am designated the lead of
> two (cvmi and icedtea), yet I'm not a member of any group, including
> the group of OpenJDK members.  So presumably I have a history of
> contributions which is enough to be a reviewer (the role is defined as
> 'an experienced Committer') but not enough to make me a member of
> OpenJDK.  That seems odd.  This was my main worry when these bylaws
> were proposed; allowing external contributors into the fold.

It is most definitely the intent to allow external contributors into the
fold.

> Even though a member is defined as 'a Contributor who has demonstrated
> a history of significant contributions to the Community', the
> automated creation of members is based on the current group system,
> not contributions.

Right.  The transition plan [3] was deliberately written to use current
Group Membership data rather than mandate that someone try to discern who
should belong to which Group based upon past contributions.  The latter
approach could too easily lead to controversial decisions.  Insofar as
there are people -- including you -- who pretty obviously deserve to
belong to various Groups, I'll be encouraging appropriate Group Leads
to consider nominating them shortly after the Bylaws go into effect.

>                     It's not clear to me how group membership will
> grow or how new groups will be created, given the current membership.
> I know the process (OpenJDK members create groups, group members
> nominate new members), but I haven't seen many people being voted into
> groups in the past.

Under the old informal guidelines there wasn't much benefit to being a
Group Member, so my best guess is that people just weren't motivated.

> ...
> 
> Porters Group
> =============
> This follows on from the previous point about electing new members.
> 
> This group seems to be sponsoring seven projects (bsd, caciocavallo,
> haiku, icedtea, macosx, mips and zero) yet none of the leads of these
> projects are members of the group.  Indeed, the group only has three
> members, including the lead.

There's no requirement that the Leads of the Projects being sponsored by
a Group be Members of that Group.

> I'd also see IcedTea as being a distinct group rather than a porting
> effort,

It was originally proposed as a Project, so that's what it is at the
moment.

>         but obviously no-one working on IcedTea is eligible to propose
> such a group.

I'm confident that will change fairly soon.

> jdk7 & jdk8
> ===========
> I've been given a role as reviewer for both these projects (thanks for
> that).  However, I'm not clear how this works.  I've never committed
> to either and I doubt many others have either, as changes are fed
> from other repositories (tl, awt, 2d, build, etc.) into these trees
> by the release team at Oracle.  So how will I actually review changes?
> Is there a further change planned?
> 
> I have committed to 2d, awt, tl, build and swing in the past, which I
> believe should make me a committer on these projects.  But I have no
> such role.  Indeed, I'm not even an author.  Could someone explain the
> logic here?

Both the JDK 7 and JDK 8 Projects have a set of integration forests
(tl, awt, 2d, build, etc.) into which Committers push their changes.
If you've committed to one of thoese integration forests then you have,
effectively, committed to the corresponding Project even though you
didn't push directly into the Project's master forest.

There are Groups whose names are the same as some of these integration
forests (e.g., AWT, Build, Swing), but there is no formal connection
between Groups and integration forests.

> HotSpot
> =======
> I've contributed a number of fixes to HotSpot.  However, it is impossible
> for me to become a contributor as:
> 
> 'If you have directly pushed one or more changegroups into a Project's
> Mercurial repositories then you are considered a Committer.'
> 
> Every time I have submitted a patch, I have been told this needs to go
> through an internal JPRT system and has to be done by a member of Oracle.
> So, even if I had contributed over a hundred patches, I could still not
> become a HotSpot committer.  However, someone committing a single patch
> inside Oracle can.
> 
> Was some attempt made to accomodate this inequality during the census?
> And will this requirement be removed going forward?

Sorry, but no attempt was made to address this issue during the Census.
Properly addressing it requires making the JPRT system, or something
similar, available externally.  We want to make that happen in due
course, but in the meantime I'll encourage the HotSpot Express Project
Lead to add you and other regular HotSpot contributors as Authors.

> jdk6 members
> ============
> 
> The jdk6 project membership seems a little odd to me and I was wondering
> how the positions were arrived at.  The following are all listed as reviewers:
> 
> alanb 	       Alan Bateman
> andrew 	       Andrew John Hughes
> anthony        Anthony Petrov
> chegar 	       Chris Hegarty
> darcy 	       Joe Darcy
> dcubed 	       Daniel D. Daugherty
> igor 	       Igor Nekrestyanov
> jjg 	       Jonathan Gibbons
> jjh 	       Jim Holmlund
> katleman       David Katleman
> ksrini 	       Kumar Srinivasan
> malenkov       Sergey Malenkov
> martin 	       Martin Buchholz
> michaelm       Michael McMahon
> mullan 	       Sean Mullan
> prr 	       Phil Race
> tbell 	       Tim Bell
> weijun 	       Max Weijun Wang
> 
> While I know I've committed a lot to OpenJDK6 (including several big
> HotSpot merges), I'm not aware of any others on this list having
> 'pushed at least 32 changegroups into such a Project's Mercurial
> repositories' (apologies if I'm wrong here).  I know some like
> Jonathan, Daniel and Martin have actively contributed changesets, but
> I can't recall it being more than ten.  Obviously, Joe is the
> former maintainer so he deserves the merit for that instead.
> The others I struggle to remember being involved at all.

Going strictly by the 32-changegroup threshold, the list of Reviewers for
the JDK 6 Project would be just:

    andrew  Andrew John Hughes
     darcy  Joe Darcy
       jjg  Jonathan Gibbons
     ohair  Kelly O'Hair

That seemed unnecessarily limiting, especially in light of the fact that
many contributors into JDK 6 already have the Reviewer role in JDK 7 and
its successor Projects.  In order to avoid having to vote a bunch of
people into the Reviewer role for JDK 6 just to get their backporting
work done I decided to make any JDK 6 Committer who's already a JDK 7
Reviewer into a Reviewer for JDK 6 as well.

It's been suggested to me by others that even that's too limiting, and
that instead the roles for JDK 6 should be initialized exactly as for the
JDK 7, JDK 7 Updates, and JDK 8 Projects.  I'd appreciate feedback from
you, Kelly, and Joe as to whether you think that's a reasonable approach.

> The committers are listed as:
> 
> amenkov        Alex Menkov
> aph 	       Andrew Haley
> asaha 	       Abhijit Saha
> coffeys        Sean Coffey
> gbenson        Gary Benson
> jgodinez       Jennifer Godinez
> kalli 	       Karl Helgason
> langel 	       Lillian Angel
> omajid 	       Omair Majid
> ptisnovs       Pavel Tisnovsky
> robilad        Dalibor Topić
> tdv 	       Dmitri Trembovetski
> valeriep       Valerie Peng
> 
> I know aph, gbenson, kalli, langel, omajid and ptisnovs have all
> committed to OpenJDK6 at some point.  I would say most of them have
> committed more or the same as most of those in the list above.

Yes, they have.  FYI here's the raw data:

      2 alanb
      4 amenkov
     95 andrew
      1 anthony
     10 aph
     22 asaha
      2 chegar
      2 coffeys
     44 darcy
      9 dcubed
      2 gbenson
      1 igor
      1 jgodinez
     38 jjg
      2 jjh
      6 kalli
      6 katleman
      1 ksrini
      2 langel
      1 malenkov
     13 martin
      1 michaelm
      2 mullan
    146 ohair
     10 omajid
      1 prr
     13 ptisnovs
     13 robilad
      2 tbell
      2 tdv
      1 valeriep
      4 weijun

> Again, I'm not familiar with the contributions of the others.
> 
> Can you give some insight into the data that was used for this list?

The initial lists of Committers were computed by an automated analysis of
all the Mercurial push messages found in the OpenJDK e-mail archives.

> It may be correct, but it contradicts my intuitive view of this project.
> You should be aware that most of the changesets are OpenJDK7 backports,
> so in many cases they  will have been written by one person for OpenJDK7
> and then backported by someone else, while retaining the same authorship
> as the original changeset.

Right.  That's one reason the Bylaws maintains the distinction between
changeset Author and changegroup Committer.

> On a more important point, as I'm now a reviewer for OpenJDK6, does this
> mean I can approve patches going in without also having to have them
> approved by Kelly?  That would definitely speed up work on OpenJDK6.

That's up to Kelly.  I'll let him speak for himself.

- Mark


[1] http://openjdk.java.net/groups/gb/bylaws/draft-openjdk-bylaws-10
[2] http://openjdk.java.net/groups/
[3] http://openjdk.java.net/groups/gb/bylaws/draft-openjdk-bylaws-10#_B



More information about the discuss mailing list