You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cocoon.apache.org by Reinhard Poetz <re...@apache.org> on 2004/11/29 19:37:19 UTC

Kernel integration in 2.2?

Pier Fumagalli wrote:
> On 29 Nov 2004, at 17:32, Reinhard Poetz wrote:
> 
>> Pier Fumagalli wrote:
>>
>>> On 23 Nov 2004, at 10:58, Reinhard Poetz wrote:
>>>
>>>> Pier,
>>>>
>>>> IIUC a block declares in a descriptor file all components that it 
>>>> wants to provide to other blocks by their interfaces. External 
>>>> blocks can then lookup those components because they are provided by 
>>>> the classloader.
>>>>
>>>> Is it still true that a block can have a "block-public.jar" and a 
>>>> "block-private.jar"? The block-private.jar is shielded and only 
>>>> available within the block and the block-public.jar contains all 
>>>> classes (including the interfaces of publicly available components + 
>>>> other public available classes) that will be publicly available?
>>>>
>>>> I'm  asking because the whiteboard/block-builder is based on a 
>>>> separation of public and private classes (--> JARs). This ensures 
>>>> that block A, that depends on block B, is compiled only by using 
>>>> block B's public JAR. (... and the eclipse .classpath also contains 
>>>> only block B's public JAR).
>>>
>>> Sorry, been on holiday for the last couple of weeks...
>>> No, in the new version of the kernel, all JARs are "public"... I 
>>> simply noticed that "private" classes to a block were becoming (after 
>>> writing 20 or so blocks) a major pain in the a**... And at the end of 
>>> the day, I found them quite counterproductive.
>>> Developing I found myself in the position of moving classes from 
>>> "private" to "public" when extending blocks, and ending up with 
>>> having nothing private and everything public most of the times...
>>> So, to keep the story short, no, there's no more difference, 
>>> everything is public.
>>> There is (though) an implied "local.jar" that doesn't need to be 
>>> declared in the deployment descriptor but is simply a jar file that 
>>> (if it exists) gets loaded... Look at the build script, it'll create 
>>> this jar for each block containing sources and the kernel will use it 
>>> even if not declared...
>>>     Pier
>>
>>
>> Thank you.
>> Yesterday, Sylvain and I talked on ICQ about block implementation but 
>> without considering requirements coming from the kernel 
>> implementation. I summarized it and will put it into our wiki as soon 
>> as it is finished.
> 
> 
> Thanks,
>     as I said, I've been on holiday for 2 weeks (Egypt, no internet 
> access) and I'm having a hard time catching up... Any summary will be 
> good cuz I'm running through thousands of emails in one go and the brain 
> is bubbling away! :-P
> 
>     Pier


find the summary at http://wiki.apache.org/cocoon/22BlockImplementation
--> any feedback would be *very* appreciated to get a clear todo list on what we 
have to change/implement at the existing  Cocoon core implementation in trunk

-- 
Reinhard

Re: Kernel integration in 2.2?

Posted by Reinhard Poetz <re...@apache.org>.
Stefano Mazzocchi wrote:

> Only one thing is missing: you need a "linkrewriter" in place because 
> one block does not now where the other block's URLs are mounted.


Thanks. Added to the Wiki page.

-- 
Reinhard

Re: Kernel integration in 2.2?

Posted by Stefano Mazzocchi <st...@apache.org>.
Reinhard Poetz wrote:
> Pier Fumagalli wrote:
> 
>> On 29 Nov 2004, at 17:32, Reinhard Poetz wrote:
>>
>>> Pier Fumagalli wrote:
>>>
>>>> On 23 Nov 2004, at 10:58, Reinhard Poetz wrote:
>>>>
>>>>> Pier,
>>>>>
>>>>> IIUC a block declares in a descriptor file all components that it 
>>>>> wants to provide to other blocks by their interfaces. External 
>>>>> blocks can then lookup those components because they are provided 
>>>>> by the classloader.
>>>>>
>>>>> Is it still true that a block can have a "block-public.jar" and a 
>>>>> "block-private.jar"? The block-private.jar is shielded and only 
>>>>> available within the block and the block-public.jar contains all 
>>>>> classes (including the interfaces of publicly available components 
>>>>> + other public available classes) that will be publicly available?
>>>>>
>>>>> I'm  asking because the whiteboard/block-builder is based on a 
>>>>> separation of public and private classes (--> JARs). This ensures 
>>>>> that block A, that depends on block B, is compiled only by using 
>>>>> block B's public JAR. (... and the eclipse .classpath also contains 
>>>>> only block B's public JAR).
>>>>
>>>>
>>>> Sorry, been on holiday for the last couple of weeks...
>>>> No, in the new version of the kernel, all JARs are "public"... I 
>>>> simply noticed that "private" classes to a block were becoming 
>>>> (after writing 20 or so blocks) a major pain in the a**... And at 
>>>> the end of the day, I found them quite counterproductive.
>>>> Developing I found myself in the position of moving classes from 
>>>> "private" to "public" when extending blocks, and ending up with 
>>>> having nothing private and everything public most of the times...
>>>> So, to keep the story short, no, there's no more difference, 
>>>> everything is public.
>>>> There is (though) an implied "local.jar" that doesn't need to be 
>>>> declared in the deployment descriptor but is simply a jar file that 
>>>> (if it exists) gets loaded... Look at the build script, it'll create 
>>>> this jar for each block containing sources and the kernel will use 
>>>> it even if not declared...
>>>>     Pier
>>>
>>>
>>>
>>> Thank you.
>>> Yesterday, Sylvain and I talked on ICQ about block implementation but 
>>> without considering requirements coming from the kernel 
>>> implementation. I summarized it and will put it into our wiki as soon 
>>> as it is finished.
>>
>>
>>
>> Thanks,
>>     as I said, I've been on holiday for 2 weeks (Egypt, no internet 
>> access) and I'm having a hard time catching up... Any summary will be 
>> good cuz I'm running through thousands of emails in one go and the 
>> brain is bubbling away! :-P
>>
>>     Pier
> 
> 
> 
> find the summary at http://wiki.apache.org/cocoon/22BlockImplementation
> --> any feedback would be *very* appreciated to get a clear todo list on 
> what we have to change/implement at the existing  Cocoon core 
> implementation in trunk

Only one thing is missing: you need a "linkrewriter" in place because 
one block does not now where the other block's URLs are mounted.

-- 
Stefano.