In that central xml, we can refer to multiple versions of the same
feature descriptor:

<features>
  <repository>mvn:org.apache.camel/camel-karaf/xml/features/2.7.0</repository>
  <repository>mvn:org.apache.camel/camel-karaf/xml/features/2.8.0</repository>
 ...
</features>

I think we'd need a descriptor for each minor version of Karaf so that
we can make sure only features that have been tested on a given karaf
version are available.

On Wed, May 4, 2011 at 13:28, Christian Schneider
<[email protected]> wrote:
> I also think a small karaf with the easy possibility to create custom
> distros is the way to go.
> A central list of pointers to repository files makes sense. But we have to
> do this a bit different than the current feature files. Currently a url to a
> feature file always points to a certain version of that file. For a central
> list this does not make sense.
>
> So I think we rather need a list of the base urls without version and then
> an easy way for users to install a feature file with a certain version. So
> for example to install the feature url for camel the user should be able to
> write something like:
> features:addurl camel 2.7.0
>
> Do we already have support for this or something similar? Or do we have an
> issue for it? If not I can create one.
>
> Christian
>
>
> Am 04.05.2011 13:21, schrieb Guillaume Nodet:
>>
>> I think we need a way to enable user to install other features easily
>> without having to release karaf for that.
>> It just does not scale if we have to release Karaf because Camel as
>> released a new version for example.
>> We've already discussed that some time ago and I think we need to find
>> a good technical solution for that.
>> Maybe having a xml feature descriptor referenced at
>> http://karaf.apache.org/features/repository.xml which would point to
>> various other repositories (such as camel, cxf, servicemix, web,
>> aries, etc...) is more scalable as we would not have to release a new
>> karaf container each time one of those things change.  People may want
>> Apache Direction, OpenEJB, ActiveMQ, etc..., we can't host all those
>> things in Karaf trunk as this would create unnecessary ties between
>> the projects and Karaf.
>>
>> Once we have that, we should keep Karaf main distribution clean and
>> lean and provide all the optional bits using this way.   Combined with
>> an easy way to create custom distribution, I do think that's the way
>> to go.
>>
>> On Wed, May 4, 2011 at 13:12, Ioannis Canellos<[email protected]>  wrote:
>>>>
>>>> I think that's what we are working on already as part of 3.0, so not
>>>> sure if I really understand what you mean here.
>>>>
>>> I see clustering to be part of the core karaf distribution. By that I
>>> mean
>>> that the clustering solution should be provided as a feature inside the
>>> standard feature repository.
>>>
>>> --
>>> *Ioannis Canellos*
>>> *
>>>  http://iocanel.blogspot.com
>>>
>>> Apache Karaf<http://karaf.apache.org/>  Committer&  PMC
>>> Apache ServiceMix<http://servicemix.apache.org/>    Committer
>>> *
>>>
>>
>>
>
> --
> ----
> http://www.liquid-reality.de
>
>



-- 
Cheers,
Guillaume Nodet
------------------------
Blog: http://gnodet.blogspot.com/
------------------------
Open Source SOA
http://fusesource.com

Connect at CamelOne May 24-26
The Open Source Integration Conference
http://camelone.com/

Reply via email to