activemq-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Martyn Taylor <mtay...@redhat.com>
Subject Re: [DISCUSS} HornetQ & ActiveMQ's next generation
Date Mon, 30 Mar 2015 15:39:49 GMT
As the release manager of the initial release of ActiveMQ-6 I have kept 
up with this thread in hope that consensus would eventually be reached. 
Myself and the other folks actively contributing to ActiveMQ-6 would 
then be able to make the required changes and move forward. I'd really 
rather not kick up any more turf on this matter, but I'll just say that 
the folks working on HornetQ sent the original code grant email in good 
faith, we thought that by coming together we could achieve much more, 
we'd bring together not only two great pieces of software but also 
knowledge, experience and also, importantly, time and effort.

In the original donation thread it was decided that ActiveMQ-6 would be 
the name for the project repository housing the code donation. This was 
not forced nor was it even suggested by the HornetQ developers, we just 
accepted the state of play and started work on getting things in place. 
Given the ActiveMQ-6 name for the repo myself and others (perhaps 
incorrectly) assumed that ActiveMQ 6.0.0 followed. There is nothing more 
to the initial release proposal than that, this is just how we 
understood what had been agreed. There is no hostile take over attempt 
by any ActiveMQ-6 developers or any corporate bullish strategy that 
others have implied on this thread.

I'd like to also say that myself and the other contributors to the 
ActiveMQ-6 project were very excited about the response to initial code 
grant. We took the bull by the horns and really put blood, sweat and 
tears into getting existing code base in shape. The last 5 months have 
been extremely hard work and we have achieved a great deal, I am 
extremely excited to get an inital release out in the open for people to 
evaluate and express their thoughts and views of the project. I'd love 
to make progress and finish what we had started in getting an initial 
release of the code base out in the open. It appears that there are 
others also excited about trying it out. So, let's pull together and get 
this thing out there.

 From what I have read thus far, that there are individuals in the 
community that that have very different view points, and opinions are 
not likely to be swayed completely to one side or the other. This is 
completely understandable and I can empathise.

Personally, I have no affiliation to ActiveMQ 6.0.0 other than this is 
what I had understood had been agreed. To me, it also follows (and I 
realise that other have objected to this) that this would mean 
ActiveMQ-6 would become the next generation of the ActiveMQ broker 
(Given that we are on the 5.x series now). There was obviously 
confusion, perhaps misunderstanding here.


I wonder, where do we go from here?  There have been a couple of 
suggestions thus far that attempt to address some of the issues that 
have arisen in the thread:

1. Use a different project name akin to Apollo.  This addresses some of 
the concerns that ActiveMQ-6 should not be positioned as the next 
generation of the existing broker.  But also implies that the projects 
are still distinct, separate things, which perhaps, detracts from the 
original goal of the code donation email, which was an offer to join 
forces to create a unified, single great broker.  In addition, from what 
I have read from Chris Mattmann, it appears that ASF are not keen on 
creating sub projects, which may end up rendering option 1. -> 3.

2. Put some clear daylight between the existing ActiveMQ broker and the 
version of the next initial release of the ActiveMQ-6 code base, whilst 
using milestone releases e.g. 10.0.0-M1.  This approach addresses some 
of the concerns that ActiveMQ could never do another major release but 
work could continue on both code bases.  However, this still implies 
that the ActiveMQ-6 core, would replace the existing core at some 
undecided point in the future.

3. Move the ActiveMQ-6 into incubator as a separate project.  This does 
address the concerns of those that are against using the ActiveMQ-6 code 
base as the core.  But seems to completely detract from the original 
goal of joining forces and creating a "one broker to rule them all" 
taking the best parts of ActiveMQ-6 (the core) and adding all the cool 
features from the existing ActiveMQ code base.

Perhaps there are other suggestion that I have missed.

To me, approaches 1. and 3. seem like they will inevitably result in 
ActiveMQ-6 becoming a separate, distinct project.  Which goes against 
what we were really trying to achieve here, which was to unite and 
combine forces, this was the basic intent of the original code grant 
email.  It would, in my opinion be a tragedy, if after the initial 
prospect of collaboration, months of hard work and the prospect of 
bringing the best pieces of both worlds, to create something even 
better, the projects parted and went their separate ways...

In an attempt to pull things together and make progress.  How would 
people feel if I proposed a release candidate based on suggestion 2. 
i.e. ActiveMQ 10.0.0-M1.  We could use this "10 series" to move forward 
with combining the ActiveMQ existing features with the fast ActiveMQ-6 
core and address migration.  We could then decide as a community if and 
when a 10.0.0 release is in a suitable state to be release as the next 
generation?

If people are not on board with this, what would you suggest?  How can 
we move forward?

Regards
Martyn

On 30/03/15 05:23, David Jencks wrote:
> Hi Art,
>
> Thanks for trying to bring me back to earth :-)
>
> I think I understand a bit more of what you are concerned about, and your concerns are
definitely worth discussing, although I think in some of your earlier posts we disagree a
lot on what is going on.
>
> - name:   I don't think anyone cares any more that formerly-hornetQ be called activemq-6
right now.  I hope this would alleviate your concern that it will necessarily be the next
activemq even if it doesn't work :-)
>
> - need for new broker:  Other people have explained way better than I can why a new broker
might greatly broaden where activemq could be used.  I don't want to see activemq disadvantaged
on say relatively slow processors with a lot of cores.
>
> - concern about backwards compatibility and getting into a no-migration bind.  This is
a big problem, and a big danger, and Raj seems to be saying it could take years to make it
completely backwards compatible (hopefully not replicating bugs :-)  However, if I understood
his post correctly, activemq has already had 4 broker replacements and it's still going strong.
 So I don't see this as an insuperable obstacle.
>
> You've also said some things that don't match up with what it looks to me like is going
on.  I'm pretty sure you are more involved than I, so you might have more evidence, but I
haven't seen it.
>
> - hornetQ is replacing activemq, rather than merging code into the existing activemq
code base.  I've tried to address this repeatedly.  The only way I can imagine the integration
working, since everything is attached to the broker, is to start with the new broker and add
everything that isn't the broker to it, changing both as needed so it works.  As far as I
can tell this is exactly what is happening.  What other plausible merge/integration strategy
can you imagine?  At the beginning of this process the new code repository is going to look
like former-hornetq with the name changed.  As bits get added it's going to look more and
more like activemq 5 does now.
>
> - hornetQ is going to continue to exist as a separate messaging solution and entity.
 My understanding has always been 100% that the hornetQ intent is to merge the code bases,
drop the hornetQ name, and not have any separate hornetQ code base, community, project, product…..
>
> Finally, I don't see what the board can offer for these questions, I think the community
has to decide what it wants to do.  There' might well be community problems and I'd expect
the board to address those.  Unless I'm wrong about the last point, hornetQ continuing as
a separate project, I don't see any of these as community problems but rather technical decisions
about the project direction.
>
> I sure hope we can continue with more communication and less noise :-)  And I hope I
haven't missed any concerns you regard as important.
>
> thanks
> david jencks
>
> On Mar 28, 2015, at 2:24 PM, artnaseef <art@artnaseef.com> wrote:
>
>> David - please go back and read my posts (user name artnaseef, full name
>> Arthur Naseef).  I have repeated myself multiple times with concerns.  And
>> there has not been constructive response to my concerns, nor to questions I
>> posed in an attempt to get clarity on the position that ActiveMQ needs a new
>> broker.
>>
>> It is disappointing because I know there is valid discussion there.
>>
>> I agree this thread contains much passion and input that is unactionable
>> (i.e. pure criticism), and that sucks because it will never serve to move
>> use forward, reach conclusion, nor build consensus.  At the same time, it's
>> understandable and I recognize that I have inserted some myself.  So let me
>> be the first to apologize.  I'm sorry for statements that I've made which
>> have not been constructive.
>>
>> Getting back to the actionable concerns raised and finding a way to address
>> them going forward would be greatly appreciated.
>>
>> If you want me to rehash my concerns, then I'll do so, but I would prefer to
>> avoid repeating myself multiple times.
>>
>>
>>
>> --
>> View this message in context: http://activemq.2283324.n4.nabble.com/DISCUSS-HornetQ-ActiveMQ-s-next-generation-tp4693781p4694024.html
>> Sent from the ActiveMQ - Dev mailing list archive at Nabble.com.


Mime
View raw message