tez-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hitesh Shah <hit...@apache.org>
Subject Re: Tez graduation [Was: Request for mentor assessment]
Date Mon, 23 Jun 2014 20:20:27 GMT
Hi folks, 

To be very clear, the project's source of truth has always been its status file: https://svn.apache.org/repos/asf/incubator/public/trunk/content/projects/tez.xml

For the tez website, http://tez.incubator.apache.org/team-list.html, the onus was for the
committers to test out their commit privileges, update the teamlist in the source tree and
publish the updated website. I understand it has created a lot of confusion for folks outside
of the project. I will go ahead and update the internal list to match the source of truth.

Even though I am well aware that some of the initial committers on the incubator proposal
have not been active ( my take on active is as simple as sending a single mail to any mailing
list ) on the project since inception, my understanding had been that all committers and mentors
from an incubator podling become PMC for the top level project when it graduates. Is my understanding
incorrect? In any case, to directly address the diversity question, I believe the PMC would
have members from companies such as Hortonworks, Yahoo, LinkedIn, Twitter to name a few.

As for our community, we have had multiple talks and meetups both on Tez alone as well as
ones with the folks from Hive and Pig. 
  - http://www.meetup.com/Apache-Tez-User-Group. The only meetup we have had till date had
a detailed talk on Tez in addition to presentations from folks from Hive and Pig on how they
were using or planning to use Tez.
  - There was a whole evening devoted to Tez and applications on Tez as part of the Bay Area
Hadoop User Group - http://www.meetup.com/hadoop/events/116895522/
  - Talks - there have been quite a few talks on it over the past year. For example, one can
look at the recently completed Hadoop Summit in San Jose and search for Tez related talks
of which only one talk was just about Tez and the others about applications built on top of

We have been working with the Hive and Pig folks mostly via JIRA to address their requirements
as they built out their platforms to work on Tez. One can look at https://issues.apache.org/jira/browse/HIVE-4660
and https://issues.apache.org/jira/browse/PIG-3446 to see the considerable amount of work
both these communities have put in to get their respective applications to work on top of
Tez. Most of the interaction has been on JIRA and not really on the dev mailing lists. Most
design discussions also tend to happen there. Subscribing to issues@ might give one a better
perspective on community interaction.

To be frank, the Tez community for the past year has been working towards helping make Apache
Hive and Apache Pig faster. As the platform evolves and grows, other applications such as
Flink are considering it too. Hopefully, at some point down the line, we would like to work
with the MapReduce community to have them consider using Tez. 

— Hitesh

On Jun 23, 2014, at 12:50 AM, Ted Dunning <ted.dunning@gmail.com> wrote:

> Let me repeat that my desired outcome here is Tez graduation.
> It is just that I want to see Tez graduate as a viable project.
> On Mon, Jun 23, 2014 at 12:43 AM, Mattmann, Chris A (3980) <
> chris.a.mattmann@jpl.nasa.gov> wrote:
>> Thanks Ted, these are all important points that require responses. I have
>> CC'ed the dev@tez.i.a.o
>> list, where I hope the Tez community can respond to your points below in
>> detail.
>> Guys, please see Ted's comments below and please try to address them in
>> light of the
>> recent [DISCUSS] Graduation thread I raised. It's fine if the answer to
>> some of the
>> points is (1) we plan on addressing them by X date, with Y action; (2) we
>> don't think
>> this is a valid point *because* ..*explanation*; (3) you didn't consider
>> this factoid
>> _here_, etc.
>> I'm hoping the other mentors on the project can step up and have some
>> insight
>> into the below. I look forward to the discussion.
>> Cheers,
>> Chris
>> -----Original Message-----
>> From: Ted Dunning <ted.dunning@gmail.com>
>> Reply-To: "general@incubator.apache.org" <general@incubator.apache.org>
>> Date: Monday, June 23, 2014 12:12 AM
>> To: "general@incubator.apache.org" <general@incubator.apache.org>
>> Subject: Re: Tez graduation [Was: Request for mentor assessment]
>>> On Sun, Jun 22, 2014 at 9:33 AM, Mattmann, Chris A (3980) <
>>> chris.a.mattmann@jpl.nasa.gov> wrote:
>>>>> Regardless of whether it is an explicit requirement to have some
>>>> specific
>>>>> level of diversity, there is a requirement that the group demonstrate
>>>>> community building.
>>>> I'm also sympathetic to this: you can see my thread here where
>>>> originally
>>>> I didn't
>>>> think much had been going on until I explicitly checked the lists and
>>>> went
>>>> through
>>>> the archives and saw ~7 IIRC PPMC members/committers added since
>>>> entering
>>>> [1]. That's
>>>> community building regardless of what affiliation those new members came
>>>> from.
>>>> The other metrics and statistics (health of the mailing lists/open
>>>> discussions have
>>>> been increasing; releases have been made, etc.) are all there and
>>>> outlined
>>>> in [1].
>>> As I read the private list, there has been absolutely no effort at adding
>>> committers until last month when the topic of graduation has come up.
>>> There
>>> is no substantive discussion on the list of the potential committers, but
>>> simply rather perfunctory +1 votes.
>>> In reading the thread about whether there should be/is a diversity
>>> requirement, I come to a bit of a different conclusion as well.  What I
>>> see
>>> in the discussion is that the participants seem to agree that
>>> - there should not be a simple count-based diversity measure because this
>>> would dissuade valuable commercial contributions
>>> - the chance/likelihood of project failure due to a dominant commercial
>>> contributor pulling out should not be considered since project dormancy or
>>> retirement isn't a bad thing.  Better to try and fail than never try.
>>> - there is clear recognition of the risk of commercial companies inventing
>>> a project not so much because of a desire to build community but rather as
>>> a marketing maneuver.
>>> - there is no mention of the problems that have arisen in the Hadoop
>>> groups
>>> relative to use of groups and Apache trademarks for marketing purposes.
>>> I don't want to impute motives to actions since I cannot know what others
>>> hold in their hearts, but I don't see evidence of efforts to really build
>>> the community.  There was one meat-space meetup a year ago.  There have
>>> been a few presentations of what Tez is.
>>> I do see a fair number of public statements of the form "Tez is
>>> Hortonworks
>>> answer to Impala" or "The future of Hadoop runs on Tez", most of which
>>> seem
>>> to come out of Hortonworks' marketing department.
>>> In looking at the dev list back to September of last year, I see a large
>>> number of procedural messages (votes, results, release mechanics), a few
>>> user questions (NPE in such and so, how do I install) and very few design
>>> discussions.  There was one thread last November about combiners, another
>>> in January about map parallelism, and one more in March about reducer
>>> parallelism.  To me this looks a lot like a project where all design
>>> discussions are occurring off-list.
>>> What I would like to see would include
>>> - a strong effort on the part of Tez to bring all design and
>>> implementation
>>> discussions to the list,
>>> - continuous community engagement efforts such as weekly hangouts in
>>> different timezones to help new contributors understand what is happening
>>> technically in the project in an interactive setting
>>> - lots of public talks focussed on how outsiders can contribute and how
>>> the
>>> design works
>>> - cross posting on related projects offering cross pollination
>>> opportunities.  Such related groups might be Drill, Optiq, Tajo and Spark.
>>> These postings would say things like "how could you guys help us with
>>> Tez"
>>> or "Tez has this and such, would that be helpful to you guys".
>>> - inviting contributors from related projects to give technical
>>> presentations in the context of Tez
>>> Now, my searches have been fairly cursory and could have missed important
>>> activities, but I have a bit of a hard time believing that I have missed
>>> major efforts along these lines.
>>> Based on this evidence, it really does look to me like Tez is a purely
>>> commercial development masquerading as an Apache community.
>>> I agree that superficial diversity metrics are counter-productive, but I
>>> also assert that there isn't any obvious evidence of serious community
>>> building here and there is significant evidence that building community
>>> isn't even the point of the project.  That *is* a substantive issue
>>> relative to graduation.
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
>> For additional commands, e-mail: general-help@incubator.apache.org

View raw message