[jsr294-modularity-eg] Simple Module System Proposal

[Peter Kriens]

Though Jigsaw is not unlikely to be the RI, I think an RI would  
improve a tad if there was a specification to implement. Trying to  
"cook" this specification is what you're witnessing. Unfortunately,  
this is a messy process where there are no clear design documents  
because we try to find the common ground in the EG.

SMS is a proposal from BJ, Richard and me to break the log-jam we  
experienced in the JSR 294 group concerning module membership. The  
current phase is the haggling and doing the negotiations necessary to  
make it acceptable to all. If we can find common ground about the  
concepts, then our next task will be to put this design on paper.

So please bear with us and allow the EG to do its work ...

Kind regards,

	Peter Kriens








On 17 sep 2009, at 15:51, Stephen J. McConnell wrote:

> On Thu, 2009-09-17 at 09:11 -0400, BJ Hargrave wrote:
>> Stuart is correct. I am objecting to all SMS modules in one class
>> loader. The SMS proposal suggests a 1-1 mapping of module to class
>> loader.
>
> OK - you getting me a little confused here. It seems to me as an
> observer, that SMS is an acronym initiated by Peter as an abbreviation
> for the suggestion that the expert group consider a "Simple Module
> System Proposal". I should also note that I have not seen any
> indications of consensus or agreement that the proposal is anything  
> more
> than proposal to pursue some noble aims and objectives.  Clearly,  
> SMS is
> undefined and remains a topic for you, Peter and Richard put some
> clothes around.
>
> On the other-hand - if you are using SMS as a generalised reference to
> any potential implementation of the evolving 294 modularisation work,
> then I still have a problem, because when I look at the proposed
> language changes [1] I see a description of modularisation data and  
> the
> description of semantics that link modules with classloader machinery.
> If I drill down into the Jigsaw project code [2], I see an
> implementation framework that allows the mapping of module-info data  
> to
> Configurations that expose Context objects that aggregate Modules (via
> ModuleInfo references) into classloaders.  As I mentioned earlier,  
> that
> framework seems to be mapping modules to classloaders on a 1:1 basis
> (with the note that this can be a n:1 relationship when we take into
> account the local modifier on the requires directive).
>
> Clearly neither of these scenarios make sense.
>
> Please set me strait.
>
> Cheers, Steve.
>
>
> [1] http://openjdk.java.net/projects/jigsaw/doc/language.html
> [2] http://cr.openjdk.java.net/~mr/jigsaw/api/
>
>
> -- 
> Stephen J. McConnell
> mailto:mcconnell at dpml.net
> http://www.dpml.net
> mobile: +61 4 5800 3980
>