hadoop-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Doug Cutting <cutt...@apache.org>
Subject Re: ZooKeeper as TLP discussion results.
Date Wed, 07 Apr 2010 19:00:50 GMT
Patrick Hunt wrote:
> Please find the results of the ZooKeeper as TLP discussion here.
> http://bit.ly/c4fuZT
> There was consensus amongst the development team that we will stay as a 
> subproject of Hadoop for the time being. Full details of the discussion 
> can be found in the thread provided.

In that discussion you state:
 > a) I do not think ZooKeeper currently has a sufficiently large and
 > diverse enough community such that it can fend for itself as a
 > TLP

How does the Hadoop PMC assist Zookeeper?  I see very little evidence 
that the Hadoop PMC is involved at all day-to-day operations of 
Zookeeper.  Currently the Hadoop PMC must vote on new Zookeeper 
committers and releases.  These votes usually pass with the minimum 
required number of PMC votes, often only after an appeal is made for 
more votes.  This does not seem like significant oversight.  Beyond 
that, I see little evidence of involvement by the Hadoop PMC in 
Zookeeper.  So I don't see this as a strong argument not to become a TLP.

If a group of committers is operating independently, then they ought to 
be an independent TLP.  The fact that you're a subproject operating 
independently only hides the lack of diversity, it doesn't help it.

 > b) Loss of branding and discover-ability

This also seems a poor reason to remain a sub-project.  We can retain 
prominent links to Zookeeper from hadoop.apache.org regardless of how 
the projects are structured.

 > c) "if ain't broke don't fix it". I have frequent interactions with
 > Hadoop PMC/Chair and an Apache board member.

That should not change as a TLP.  Apache encourages cross-project 
communication and collaboration.

The things that would change if Zookeeper were a TLP are:
  1. its official website would be at zookeeper.apache.org.
  2. it could vote for committers and releases directly rather than 
through the Hadoop PMC
  3. it would submit its quarterly report to the board directly, instead 
of via the Hadoop PMC.

That's pretty much it.

The board has a well established report and review system.  Each 
project's quarterly report is closely read and must be individually 
approved by a quorum of the board's members.  PMCs tend not to have such 
a review mechanism for subprojects.  Historically subprojects that 
develop problems have been slow to identify, and the problems have 
worsened in the meantime.  The focus of each PMC should be on direct 
decisions about code, committers and releases.  The board's job is to 
make sure that each PMC operates effectively.  These are different 
responsibilities and require different processes.

No one is going to force Zookeeper to become a TLP against its will, and 
no change must be made immediately.  But I think such a move would be 
easy to make, have significant upside for the project in simplifying its 
formal votes, and have significant upside to the foundation in 
facilitating the project's direct interaction with the board.


View raw message