[DISCUSS] Replace user@f.a.o with Stackoverflow

[Ufuk Celebi-2]

Hey all,

I've been asking myself: how can we make it as easy as possible for future
users who run into problems to find existing answers?

As an example take this answer from Stephan [1]. This is super valuable
feedback and a very specific question. Ideally, a new user who runs into
the same problem should find this answer.

I think the best way to achieve this is Stack Overflow. There are already a
couple of questions regarding Flink, but they questions are fragmented over
user@ and SO.

Robert already added a suggested tag to the website recently, but I was
wondering whether we should take this a step further and want to discuss
the following:

Should we *only* recommend SO for user questions and don't link the user
list at all?

For example, the Netty project is doing it that way. This has the advantage
that 1) there is one conveniently searchable place for user questions, and
2) all the SO pros like voting on answers, it is well known etc. On the
other hand, you can already search the ML archives with Nabble and what
about existing questions/answers?

Looking forward to get your feedback. It's also quite possible that there
is no real problem here and that I'm trying to over engineer this. Feel
free to tell me. ;-)

– Ufuk

[1]
http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html

[Henry Saputra]

Unfortunately as top level Apache project we can not have official
discussions outside Apache domains.
Which means that we still need user@ list as the main mechanism for
users to ask to communicate and ask questions.

Some Apache projects like Hadoop and Spark also have the community
monitoring Stackoverflow for questions but still maintain user@ list
as the official channel for communication for the project

- Henry

On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]> wrote:

> Hey all,
>
> I've been asking myself: how can we make it as easy as possible for future
> users who run into problems to find existing answers?
>
> As an example take this answer from Stephan [1]. This is super valuable
> feedback and a very specific question. Ideally, a new user who runs into
> the same problem should find this answer.
>
> I think the best way to achieve this is Stack Overflow. There are already a
> couple of questions regarding Flink, but they questions are fragmented over
> user@ and SO.
>
> Robert already added a suggested tag to the website recently, but I was
> wondering whether we should take this a step further and want to discuss
> the following:
>
> Should we *only* recommend SO for user questions and don't link the user
> list at all?
>
> For example, the Netty project is doing it that way. This has the advantage
> that 1) there is one conveniently searchable place for user questions, and
> 2) all the SO pros like voting on answers, it is well known etc. On the
> other hand, you can already search the ML archives with Nabble and what
> about existing questions/answers?
>
> Looking forward to get your feedback. It's also quite possible that there
> is no real problem here and that I'm trying to over engineer this. Feel
> free to tell me. ;-)
>
> – Ufuk
>
> [1]
> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html

[Ufuk Celebi-2]

Thanks for the info Henry. Then I'll just add some text to the webpage to
make sure people know how to look for old ML answers via Nabble. Should do
the job as well. :-)

On Saturday, April 18, 2015, Henry Saputra <[hidden email]> wrote:

> Unfortunately as top level Apache project we can not have official
> discussions outside Apache domains.
> Which means that we still need user@ list as the main mechanism for
> users to ask to communicate and ask questions.
>
> Some Apache projects like Hadoop and Spark also have the community
> monitoring Stackoverflow for questions but still maintain user@ list
> as the official channel for communication for the project
>
> - Henry
>
> On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]
> <javascript:;>> wrote:
> > Hey all,
> >
> > I've been asking myself: how can we make it as easy as possible for
> future
> > users who run into problems to find existing answers?
> >
> > As an example take this answer from Stephan [1]. This is super valuable
> > feedback and a very specific question. Ideally, a new user who runs into
> > the same problem should find this answer.
> >
> > I think the best way to achieve this is Stack Overflow. There are
> already a
> > couple of questions regarding Flink, but they questions are fragmented
> over
> > user@ and SO.
> >
> > Robert already added a suggested tag to the website recently, but I was
> > wondering whether we should take this a step further and want to discuss
> > the following:
> >
> > Should we *only* recommend SO for user questions and don't link the user
> > list at all?
> >
> > For example, the Netty project is doing it that way. This has the
> advantage
> > that 1) there is one conveniently searchable place for user questions,
> and
> > 2) all the SO pros like voting on answers, it is well known etc. On the
> > other hand, you can already search the ML archives with Nabble and what
> > about existing questions/answers?
> >
> > Looking forward to get your feedback. It's also quite possible that there
> > is no real problem here and that I'm trying to over engineer this. Feel
> > free to tell me. ;-)
> >
> > – Ufuk
> >
> > [1]
> >
> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html
>

[Robert Metzger]

I think the mailinglists are indexed by Google as well, through various web
based mirrors.

SO is good for questions with clear answers, but if a user wants more
interactive feedback (with questions back and forth) the user@ list is
probably better.
Also, people which are experienced with Apache projects are probably
expecting a user@ list.

I think our current approach of having both will lead to the best user
experience: Everybody can use what they like the most.

Regarding the Hashjoin exception: Stephan can ask and answer the question
himself on SO. Thats a nice way of collecting some points ;)




On Sun, Apr 19, 2015 at 2:15 PM, Ufuk Celebi <[hidden email]> wrote:

> Thanks for the info Henry. Then I'll just add some text to the webpage to
> make sure people know how to look for old ML answers via Nabble. Should do
> the job as well. :-)
>
> On Saturday, April 18, 2015, Henry Saputra <[hidden email]>
> wrote:
>
> > Unfortunately as top level Apache project we can not have official
> > discussions outside Apache domains.
> > Which means that we still need user@ list as the main mechanism for
> > users to ask to communicate and ask questions.
> >
> > Some Apache projects like Hadoop and Spark also have the community
> > monitoring Stackoverflow for questions but still maintain user@ list
> > as the official channel for communication for the project
> >
> > - Henry
> >
> > On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]
> > <javascript:;>> wrote:
> > > Hey all,
> > >
> > > I've been asking myself: how can we make it as easy as possible for
> > future
> > > users who run into problems to find existing answers?
> > >
> > > As an example take this answer from Stephan [1]. This is super valuable
> > > feedback and a very specific question. Ideally, a new user who runs
> into
> > > the same problem should find this answer.
> > >
> > > I think the best way to achieve this is Stack Overflow. There are
> > already a
> > > couple of questions regarding Flink, but they questions are fragmented
> > over
> > > user@ and SO.
> > >
> > > Robert already added a suggested tag to the website recently, but I was
> > > wondering whether we should take this a step further and want to
> discuss
> > > the following:
> > >
> > > Should we *only* recommend SO for user questions and don't link the
> user
> > > list at all?
> > >
> > > For example, the Netty project is doing it that way. This has the
> > advantage
> > > that 1) there is one conveniently searchable place for user questions,
> > and
> > > 2) all the SO pros like voting on answers, it is well known etc. On the
> > > other hand, you can already search the ML archives with Nabble and what
> > > about existing questions/answers?
> > >
> > > Looking forward to get your feedback. It's also quite possible that
> there
> > > is no real problem here and that I'm trying to over engineer this. Feel
> > > free to tell me. ;-)
> > >
> > > – Ufuk
> > >
> > > [1]
> > >
> >
> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html
> >
>

[Stephan Ewen]

I like the searchability of StackOverflow better than that of Mailing List
archives, but we need to stick to Apache rules here.

I like to idea of posting answers with a lot of effort to StackOverflow.
Not sure however how it looks if we post a lot of questions and answer them
ourselves. Looks weird...

On Mon, Apr 20, 2015 at 11:12 AM, Robert Metzger <[hidden email]>
wrote:

> I think the mailinglists are indexed by Google as well, through various web
> based mirrors.
>
> SO is good for questions with clear answers, but if a user wants more
> interactive feedback (with questions back and forth) the user@ list is
> probably better.
> Also, people which are experienced with Apache projects are probably
> expecting a user@ list.
>
> I think our current approach of having both will lead to the best user
> experience: Everybody can use what they like the most.
>
> Regarding the Hashjoin exception: Stephan can ask and answer the question
> himself on SO. Thats a nice way of collecting some points ;)
>
>
>
>
> On Sun, Apr 19, 2015 at 2:15 PM, Ufuk Celebi <[hidden email]> wrote:
>
> > Thanks for the info Henry. Then I'll just add some text to the webpage to
> > make sure people know how to look for old ML answers via Nabble. Should
> do
> > the job as well. :-)
> >
> > On Saturday, April 18, 2015, Henry Saputra <[hidden email]>
> > wrote:
> >
> > > Unfortunately as top level Apache project we can not have official
> > > discussions outside Apache domains.
> > > Which means that we still need user@ list as the main mechanism for
> > > users to ask to communicate and ask questions.
> > >
> > > Some Apache projects like Hadoop and Spark also have the community
> > > monitoring Stackoverflow for questions but still maintain user@ list
> > > as the official channel for communication for the project
> > >
> > > - Henry
> > >
> > > On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]
> > > <javascript:;>> wrote:
> > > > Hey all,
> > > >
> > > > I've been asking myself: how can we make it as easy as possible for
> > > future
> > > > users who run into problems to find existing answers?
> > > >
> > > > As an example take this answer from Stephan [1]. This is super
> valuable
> > > > feedback and a very specific question. Ideally, a new user who runs
> > into
> > > > the same problem should find this answer.
> > > >
> > > > I think the best way to achieve this is Stack Overflow. There are
> > > already a
> > > > couple of questions regarding Flink, but they questions are
> fragmented
> > > over
> > > > user@ and SO.
> > > >
> > > > Robert already added a suggested tag to the website recently, but I
> was
> > > > wondering whether we should take this a step further and want to
> > discuss
> > > > the following:
> > > >
> > > > Should we *only* recommend SO for user questions and don't link the
> > user
> > > > list at all?
> > > >
> > > > For example, the Netty project is doing it that way. This has the
> > > advantage
> > > > that 1) there is one conveniently searchable place for user
> questions,
> > > and
> > > > 2) all the SO pros like voting on answers, it is well known etc. On
> the
> > > > other hand, you can already search the ML archives with Nabble and
> what
> > > > about existing questions/answers?
> > > >
> > > > Looking forward to get your feedback. It's also quite possible that
> > there
> > > > is no real problem here and that I'm trying to over engineer this.
> Feel
> > > > free to tell me. ;-)
> > > >
> > > > – Ufuk
> > > >
> > > > [1]
> > > >
> > >
> >
> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html
> > >
> >
>

[Kostas Tzoumas-2]

If we clearly preface with a boilerplate text and a link to the users@
question it might look less weird :-)

Perhaps we can create a stackoverflow user whose purpose is to create
questions that mirror the users@ list? Not sure how much this can be
automated though

On Mon, Apr 20, 2015 at 4:25 PM, Stephan Ewen <[hidden email]> wrote:

> I like the searchability of StackOverflow better than that of Mailing List
> archives, but we need to stick to Apache rules here.
>
> I like to idea of posting answers with a lot of effort to StackOverflow.
> Not sure however how it looks if we post a lot of questions and answer them
> ourselves. Looks weird...
>
> On Mon, Apr 20, 2015 at 11:12 AM, Robert Metzger <[hidden email]>
> wrote:
>
> > I think the mailinglists are indexed by Google as well, through various
> web
> > based mirrors.
> >
> > SO is good for questions with clear answers, but if a user wants more
> > interactive feedback (with questions back and forth) the user@ list is
> > probably better.
> > Also, people which are experienced with Apache projects are probably
> > expecting a user@ list.
> >
> > I think our current approach of having both will lead to the best user
> > experience: Everybody can use what they like the most.
> >
> > Regarding the Hashjoin exception: Stephan can ask and answer the question
> > himself on SO. Thats a nice way of collecting some points ;)
> >
> >
> >
> >
> > On Sun, Apr 19, 2015 at 2:15 PM, Ufuk Celebi <[hidden email]> wrote:
> >
> > > Thanks for the info Henry. Then I'll just add some text to the webpage
> to
> > > make sure people know how to look for old ML answers via Nabble. Should
> > do
> > > the job as well. :-)
> > >
> > > On Saturday, April 18, 2015, Henry Saputra <[hidden email]>
> > > wrote:
> > >
> > > > Unfortunately as top level Apache project we can not have official
> > > > discussions outside Apache domains.
> > > > Which means that we still need user@ list as the main mechanism for
> > > > users to ask to communicate and ask questions.
> > > >
> > > > Some Apache projects like Hadoop and Spark also have the community
> > > > monitoring Stackoverflow for questions but still maintain user@ list
> > > > as the official channel for communication for the project
> > > >
> > > > - Henry
> > > >
> > > > On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]
> > > > <javascript:;>> wrote:
> > > > > Hey all,
> > > > >
> > > > > I've been asking myself: how can we make it as easy as possible for
> > > > future
> > > > > users who run into problems to find existing answers?
> > > > >
> > > > > As an example take this answer from Stephan [1]. This is super
> > valuable
> > > > > feedback and a very specific question. Ideally, a new user who runs
> > > into
> > > > > the same problem should find this answer.
> > > > >
> > > > > I think the best way to achieve this is Stack Overflow. There are
> > > > already a
> > > > > couple of questions regarding Flink, but they questions are
> > fragmented
> > > > over
> > > > > user@ and SO.
> > > > >
> > > > > Robert already added a suggested tag to the website recently, but I
> > was
> > > > > wondering whether we should take this a step further and want to
> > > discuss
> > > > > the following:
> > > > >
> > > > > Should we *only* recommend SO for user questions and don't link the
> > > user
> > > > > list at all?
> > > > >
> > > > > For example, the Netty project is doing it that way. This has the
> > > > advantage
> > > > > that 1) there is one conveniently searchable place for user
> > questions,
> > > > and
> > > > > 2) all the SO pros like voting on answers, it is well known etc. On
> > the
> > > > > other hand, you can already search the ML archives with Nabble and
> > what
> > > > > about existing questions/answers?
> > > > >
> > > > > Looking forward to get your feedback. It's also quite possible that
> > > there
> > > > > is no real problem here and that I'm trying to over engineer this.
> > Feel
> > > > > free to tell me. ;-)
> > > > >
> > > > > – Ufuk
> > > > >
> > > > > [1]
> > > > >
> > > >
> > >
> >
> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html
> > > >
> > >
> >
>

[Ted Dunning]

Henry,

I think that the requirement is that project decisions need to be made on
an Apache controlled resource.

Not sure how many project decisions are made on the user list.  I know I
answer a ton of Mahout questions on quora.

Having more than one arena for answers and reflecting between them for key
questions and answers is a good thing for users.


On Sat, Apr 18, 2015 at 5:28 PM, Henry Saputra <[hidden email]>
wrote:

> Unfortunately as top level Apache project we can not have official
> discussions outside Apache domains.
> Which means that we still need user@ list as the main mechanism for
> users to ask to communicate and ask questions.
>
> Some Apache projects like Hadoop and Spark also have the community
> monitoring Stackoverflow for questions but still maintain user@ list
> as the official channel for communication for the project
>
> - Henry
>
> On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]> wrote:
> > Hey all,
> >
> > I've been asking myself: how can we make it as easy as possible for
> future
> > users who run into problems to find existing answers?
> >
> > As an example take this answer from Stephan [1]. This is super valuable
> > feedback and a very specific question. Ideally, a new user who runs into
> > the same problem should find this answer.
> >
> > I think the best way to achieve this is Stack Overflow. There are
> already a
> > couple of questions regarding Flink, but they questions are fragmented
> over
> > user@ and SO.
> >
> > Robert already added a suggested tag to the website recently, but I was
> > wondering whether we should take this a step further and want to discuss
> > the following:
> >
> > Should we *only* recommend SO for user questions and don't link the user
> > list at all?
> >
> > For example, the Netty project is doing it that way. This has the
> advantage
> > that 1) there is one conveniently searchable place for user questions,
> and
> > 2) all the SO pros like voting on answers, it is well known etc. On the
> > other hand, you can already search the ML archives with Nabble and what
> > about existing questions/answers?
> >
> > Looking forward to get your feedback. It's also quite possible that there
> > is no real problem here and that I'm trying to over engineer this. Feel
> > free to tell me. ;-)
> >
> > – Ufuk
> >
> > [1]
> >
> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html
>

[Henry Saputra]

@Ted, you are right, decisions should be make in the dev@ list so I
was not saying we could not have users discussions it outside of
Apache land, but should be also happening in user@ list.
So I apologize to Ufuk, Robert, and Stephan if my response implied that way.

Like @Robert has said before, some people familiar or experience
working with ASF projects before will expect questions and concerns
for users are done via user@ list.
Having said that, since user@ list is not really mandatory I
definitely love to have more than one venue to help users getting
their questions answered. That is why we also have IRC channel to help
discussions.

Adding one more venue for users to get solution answered such as in
StackOverflow is definitely right way to go.

So, coming back to Ufuk discussions, I am retracting my original
response so if the community think that it is better to just use
StackOverflow than user@ list then I think we could ask more opinions
from both dev@ and user@ list if that is the preferable way.

Apologize for the confusion I may have caused.

- Henry

On Mon, Apr 20, 2015 at 8:07 AM, Ted Dunning <[hidden email]> wrote:

> Henry,
>
> I think that the requirement is that project decisions need to be made on
> an Apache controlled resource.
>
> Not sure how many project decisions are made on the user list.  I know I
> answer a ton of Mahout questions on quora.
>
> Having more than one arena for answers and reflecting between them for key
> questions and answers is a good thing for users.
>
>
> On Sat, Apr 18, 2015 at 5:28 PM, Henry Saputra <[hidden email]>
> wrote:
>
>> Unfortunately as top level Apache project we can not have official
>> discussions outside Apache domains.
>> Which means that we still need user@ list as the main mechanism for
>> users to ask to communicate and ask questions.
>>
>> Some Apache projects like Hadoop and Spark also have the community
>> monitoring Stackoverflow for questions but still maintain user@ list
>> as the official channel for communication for the project
>>
>> - Henry
>>
>> On Sat, Apr 18, 2015 at 2:16 AM, Ufuk Celebi <[hidden email]> wrote:
>> > Hey all,
>> >
>> > I've been asking myself: how can we make it as easy as possible for
>> future
>> > users who run into problems to find existing answers?
>> >
>> > As an example take this answer from Stephan [1]. This is super valuable
>> > feedback and a very specific question. Ideally, a new user who runs into
>> > the same problem should find this answer.
>> >
>> > I think the best way to achieve this is Stack Overflow. There are
>> already a
>> > couple of questions regarding Flink, but they questions are fragmented
>> over
>> > user@ and SO.
>> >
>> > Robert already added a suggested tag to the website recently, but I was
>> > wondering whether we should take this a step further and want to discuss
>> > the following:
>> >
>> > Should we *only* recommend SO for user questions and don't link the user
>> > list at all?
>> >
>> > For example, the Netty project is doing it that way. This has the
>> advantage
>> > that 1) there is one conveniently searchable place for user questions,
>> and
>> > 2) all the SO pros like voting on answers, it is well known etc. On the
>> > other hand, you can already search the ML archives with Nabble and what
>> > about existing questions/answers?
>> >
>> > Looking forward to get your feedback. It's also quite possible that there
>> > is no real problem here and that I'm trying to over engineer this. Feel
>> > free to tell me. ;-)
>> >
>> > – Ufuk
>> >
>> > [1]
>> >
>> http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Hash-join-exceeded-exception-td1086.html
>>