jackrabbit-oak-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Felix Meschberger <fmesc...@adobe.com>
Subject Re: Location of o.a.j.mk packages
Date Fri, 30 Mar 2012 11:46:59 GMT
Hi,

Am 30.03.2012 um 07:26 schrieb Michael Dürig:

> 
> 
> On 30.3.12 9:42, Angela Schreiber wrote:
>> hi felix
>> 
>> IMO your comment wasn't noise at all.

My comment about mk package being spread in two projects was in fact noise, since this is
already part of some issue discussion pointed out by Jukka.

The rest of my original message, I still don't consider noise ;-)

>> while we did reach
>> some consensus that we want to have a separate oak-api
>> we still have some disagreement on what the oak-api should
>> actually be abstraction wise and ultimately look like.
>> 
>> my interpretation of the current discussion is as follows:
>> 
>> variant 1:
>> oak-api is basically the same as mk-api but just adds some
>> validation-magic. the tree exposed by oak-api upon creation
>> of a JCR session contains the complete JCR repository.
>> -> oak-core and oak-mk should in this scenario be located
>> in the same module as the basically expose the same
>> level of abstraction.
>> 
>> variant 2:
>> oak-api is a separate API and abstraction layer on top of
>> the mk-API. mk-API just being the storage that doesn't have
>> any knowlege of workspaces, items, node types. the oak-API
>> in this variant however, was aware of the different types.
>> the validation in oak-api implementation could rely on it's
>> knowledge of workspaces, repository-unique data (versions,
>> node types, namespace) and different types of (jcr) items.
>> -> oak-core and oak-mk are completely different components
>> and therefore should be separate modules.
>> 
>> we are still this discussing this... but your confusion on
>> the current layout, i my opinion originates from this
>> not yet being decided.
> 
> I think the original idea was more like variant 2. See whiteboard 
> snapshot at the bottom of 
> http://wiki.apache.org/jackrabbit/OakComponentStructure. However name 
> space mapping and most of node type stuff was though to go into oak-jcr.
> 
> AFAIR we didn't discuss workspaces then but from that picture it seems 
> logical to me that workspace management should also go into oak-core.

Agreed. In fact, IIRC MK is just a relatively dumb tree storage. The live of it comes from
oak-core and oak-jcr brings this into the actual JCR API world.

Regards
Felix

> 
> Michael
> 
> 
> 
>> 
>> that's my view on how the current layout actually reflects
>> the level of discussion and disagreement. we will have that
>> sorted out over the next couple of weeks.
>> 
>> kind regards
>> angela
>> 
>> On 3/29/12 5:39 PM, Felix Meschberger wrote:
>>> Hi
>>> 
>>> Ok, thanks and sorry for the noise, then.
>>> 
>>> Regards
>>> Felix
>>> 
>>> Am 29.03.2012 um 11:26 schrieb Jukka Zitting:
>>> 
>>>> Hi,
>>>> 
>>>> On Thu, Mar 29, 2012 at 5:22 PM, Felix
>>>> Meschberger<fmeschbe@adobe.com> wrote:
>>>>> So I suggest we either move the mk packages to the mk project or
>>>>> rename the mk packages in core to oak instead.
>>>> 
>>>> Yes, that's the idea [1].
>>>> 
>>>> [1]
>>>> https://issues.apache.org/jira/browse/OAK-24?focusedCommentId=13231255&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13231255
>>>> 
>>>> 
>>>> BR,
>>>> 
>>>> Jukka Zitting
>>> 


Mime
View raw message