Hi Krzysztof,
A couple of years ago, I remember a discussion with Guillaume (at
ApacheCon, Vancouver, around a couple of beers ;)), where we dealt about
doing likely what you said in ServiceMix (ServiceMix acting as a
features container). It's why we started to think about something like
Cave (as a Karaf Enterprise Features Repository).
I think your idea is interesting, and it's what I aim for Karaf Cave.
I mean, now, it's not very efficient to have non-core features
"embedded" in Karaf: it's not easy for users to update to new feature
versions without updating to a new Karaf version.
I think that non-core features should be maintained and available
outside of Karaf container itself.
We can see the following Karaf sub-projects:
- Karaf (it's what we have now, the container)
- Bundles (it's the ServiceMix Bundles "moved" as Karaf subproject)
- Features (it's the non-core features, like enterprises, activiti, etc)
- Cellar
- Cave
- EIK
- WebCosnole
Now, regarding the bundles, as it's not directly related to Karaf (it's
more OSGi generic), I wonder if it makes sense to have it in Karaf. I
did a proposal in the past to do it at Felix. Mayben we can imagine to
have a Pax project for the bundles.
For now, I would leave the bundles in ServiceMix.
ServiceMix could still provide:
- bundles and specs
- NMR layer
- a assembly leveraging and gathering features
Regards
JB
On 02/08/2014 07:53 PM, Krzysztof Sobkowiak wrote:
Hi all
I think the ServiceMix community makes a good and important job which
will be still needed. But what do you think about including the features
provided by ServiceMix in Karaf as additional enterprise (or even esb)
features (as the features for jpa provider, jdbc, jndi,....).
Karaf allows to add the Camel, ActiveMQ, CXF (and more other) features
by feature:repo-add command, but the user must still find out which
version of Camel with which version of ActiveMQ will be correctly work
with the current version of Karaf (e.g. 3.0.0). This problem is solved
in SMX which is shipped with the well defined version of each component.
Instead of having such custom distribution like SMX one would prefer to
have some features like karaf-messaging-support, karaf-services-support,
karaf-routing-support (or at least how-tos or predefined default
versions for feature:repo-add) included in Karaf which allows him to add
the routing, messaging or web service support by installing one or more
features (referencing the Camel, ActiveMQ, CXF,... features in the
proper versions and adding some extensions like the missing connection
factory in the new versions of ActiveMQ) which guarantee the compatible
versions of the components. It would be also nice to have some samples
(which currently are included in SMX) and integration tests.
I think another SMX features like Activiti support are needed not only
in ESB solutions. It could be also part of Karaf enterprise features.
The ServiceMix bundles and spec could also move to Karaf as subprojects
as they provide osgified version of enterprise libraries.
So instead of having one monolithic SMX distribution we would have more
flexible modular Karaf which could be converted into the esb solution by
installing some features or one could easily install only some of the
esb features (e.g. only web services support). It would be also easier
use another version of Camel or other component as "proposed" by Karaf.
One could say, this is one step back, because Karaf was extracted from
SMX and made as small kernel which can be used to build custom
distributions (including SMX) and we want to move SMX features back into
Karaf. But the role of Karaf as a kernel for other distributions
wouldn't be changed. Karaf even plans to have smaller distributions. It
would only extend Karaf by next group of features - esb features. We
would have one code base which have good testes features for esb
compatible with the current Karaf version. I think it should be also
easier to keep the features working with on each next Karaf release (ad
the features would be developed together with Karaf) as upgrading the
SMX distribution to the newer version of Karaf.
Cristiano writes ServiceMix has a marketing value. ServiceMix does not
need to die quickly. It could be still provided as custom distribution
based on the esb features included in Karaf... based on the newest Karaf
version. It could be once re-branded (e.g. as Karaf ESB) which could be
moved into Karaf as next Karaf distribution (alongside the apache-karaf,
apache-karaf-net, apache-karaf-minimal...).
Sorry for my chaotic ideas collected quickly during fever.
Best regards
Krzysztof
On 08.02.2014 17:49, Claus Ibsen wrote:
On Sat, Feb 8, 2014 at 4:01 PM, Jean-Baptiste Onofré <j...@nanthrax.net>
wrote:
Hi,
I fully agree with Claus and Achim.
Karaf as the core platform is a good start, that you can extend with
Fabric
and other modules (Cellar, Cave, whatever).
Yeah I think Karaf really shines as a core container, that can be
customized and tailored to your needs.
Keep Karaf like that and it will go from strength to strength.
If you want a ready to use ESB based on the same layers, you can
start with
JBoss Fuse or Talend ESB (both available in opensource and enterprise
version).
Yeah shows the versatility of Karaf that it can be tailored into
commercial products and community versions as counter parts. So when
they can do that, you can do it too. I know of companies that does
that to build their own in-house OEM platform with a customized Karaf
as base.
Regards
JB
On 02/08/2014 08:42 AM, Claus Ibsen wrote:
On Fri, Feb 7, 2014 at 7:08 PM, Cristiano Costantini
<cristiano.costant...@gmail.com> wrote:
Hi all,
as I'm waiting for Servicemix 4.6.0 to come out because it solves some
problems with the version of some bundles, I was wondering if I should
move
to Karaf (2.3.3) instead on using Servicemix as the basis for my
application.
In fact I use Spring, Pax Web, Camel and CXF, and I'll probably need
ActiveMQ if I need to implement some specific EIP. I need many
dependencies
from the servicemix bundles of wrapped dependencies, but I don't other
ServiceMix features, especially NMR that I understand from
SMX4NMR-319 is
blocking release of 4.6.0.
What you suggest me to do?
Thank you!
Cristiano
ServiceMix is a much less active project today than it used to be.
Also most of the work that used to be at SerivceMix is now happening
at Karaf and Camel instead.
Users looking for a single download installation can still find value
in ServiceMix. But its a bit concerning that the project does not do
releases so often.
In my mind ServiceMix is a dying project, and users should take that
into account.
I would point people 2 main ways.
1)
Karaf and then install what they need such as Camel / CXF etc.
Then you can build your own ServiceMix.
2)
fabric8 which is has a lot more to offer.
http://fabric8.io/
certainly for the new era of cloud and managing a lot of containers,
and having a consistent web user interface using hawtio, and much
more.
.. and besides fabric8 you may find value in Apache Karaf Cellar, and
possible other 3rd party projects.
--
Jean-Baptiste Onofré
jbono...@apache.org
http://blog.nanthrax.net
Talend - http://www.talend.com
--
Jean-Baptiste Onofré
jbono...@apache.org
http://blog.nanthrax.net
Talend - http://www.talend.com
