On Tue, Aug 16, 2011 at 2:12 PM, Martin Grigorov <[email protected]> wrote:
> Isn't option 3 the solution which we have currently in WicketStuff's
> wicket-bundle-parent ?
> If it is not then can you please give more details what exactly you mean.
>
> I'd prefer the following approach for Wicket OSGi:
> - use wicketstuff/wicket-bundle for OSGi deployments (wicket-bundle
> merges -util, -request and -core into one .jar)
> - use PAX-Wicket or wicketstuff/wicket-osgi for all glue code like
> OSGiClassResolver, OSGiInjector, serialization... And this project
> will depend on wicketstuff/wicket-bundle
>
> If any changes are needed in Wicket -core, -ioc, ... then please
> create tickets (or comment in WICKET-3737) and we will improve where
> needed.
>
> For Wicket 1.6 we could fix the split package problem in -util and
> -request and wicket-bundle may be retired
And of course PAX/wicket-osgi project could be merged to Wicket distro
if we decide so.
>
> Do you see any problems with this approach ?
>
> -util, -request and -core were made this way to improve Wicket code
> itself. Now code in -util doesn't use code from -request or -core.
>
> On Tue, Aug 16, 2011 at 10:20 AM, Brian Topping <[email protected]> wrote:
>> Option 2 kind of speaks to the idea that Martijn just had about users being 
>> less inconvenienced by fewer artifacts and it would save a lot of hassle WRT 
>> these packages split across multiple artifacts.
>>
>> We were kind of wondering out loud the other day why the packages are split 
>> across artifacts like that.  I thought I had heard the reason but was not 
>> sure I could recite it exactly.  Can anyone share?
>>
>> My non-binding vote is for @2.
>>
>> On Aug 15, 2011, at 10:56 PM, Andreas Pieber wrote:
>>
>>> Hey Guys,
>>>
>>> While we're still in RC phase we might want to consider the following point:
>>> Packaging
>>>
>>> Wicket comes in with three core artifacts:
>>>
>>> <dependency>
>>>    <groupId>org.apache.wicket</groupId>
>>>    <artifactId>wicket-core</artifactId>
>>>    <version>1.5-RC5.1</version></dependency>
>>>
>>> <dependency>
>>>    <groupId>org.apache.wicket</groupId>
>>>    <artifactId>wicket-util</artifactId>
>>>    <version>1.5-RC5.1</version></dependency>
>>>
>>> <dependency>
>>>    <groupId>org.apache.wicket</groupId>
>>>    <artifactId>wicket-request</artifactId>
>>>    <version>1.5-RC5.1</version></dependency>
>>>
>>> All three of them are required. Currently I think it is assumed best
>>> practice to include wicket-core into your maven project and let it resolve
>>> util and request automatically. Thinking about the OSGi integration into
>>> wicket this leaves us with three options:
>>>
>>>   1. OSGifying all three of them using split-package.
>>>   2. Embed util and request into wicket-core.
>>>   3. Provide e.g. "apache-wicket" which repacks core, util and request into
>>>   one package.
>>>
>>> @1: Split-Package is a real pain in OSGi. While it might work there are
>>> always some hidden risks using this technique and I would really like to
>>> avoid it at any chance
>>> @2: This one would be an option since it will be the less interrupting one.
>>> @3: This one might be the most clean solution. Among the apache-wicket
>>> artifactId we could distribute a bundle (which can be included/referenced in
>>> a classical J2EE env (as wicket-core before)), but could also be launched in
>>> an OSGi environment. In addition we can also provide e.g. features.xml &
>>> .kar packages (for Karaf) within the aritfact id, eba packages (Apache
>>> Aries) and so on. Option three has the additional advantage that we can also
>>> separate the OSGi code from the core code by packing and defining it in
>>> apache-wicket.
>>>
>>> TBH I can live with both,  @2 and @3 quite well but would slightly prefer
>>> option 3. What does other think about this?
>>>
>>> Please don't get me wrong. I don't want to be intrusive in any way or do
>>> any criticism on the current way. I can live quite well with option 2 if
>>> this is the commonly agreed solution. It's just that I want to show up the
>>> options and possibilities we have.
>>>
>>> I'm very interested in your feedback.
>>>
>>> Kind regards,
>>> Andreas
>>
>>
>
>
>
> --
> Martin Grigorov
> jWeekend
> Training, Consulting, Development
> http://jWeekend.com
>



-- 
Martin Grigorov
jWeekend
Training, Consulting, Development
http://jWeekend.com

Reply via email to