I guess I wasn't thinking so much about increments or exceptions to
the default. I was more thinking that the override would be a complete
redefine rather than a tweak. I guess I see your point a bit better
now. It might be worth working out a few use cases to see whether
tweaking the default is much better than simply redefining the
behaviour.

Thanks,

David

On 25 February 2010 13:14, Graham Charters <[email protected]> wrote:
> I think for the 'deviation' model to work, each statement needs to be
> a small incremental progression from the default.  I don't yet see how
> that would work.  If I have an "all shared" subsystem (type 1), then
> the small increment is to prevent something, whereas, if I have an
> "explicitly shared" subsystem (type 2), then the small increment is to
> allow something (would need to be a different header).  I don't think
> we should be allowing any statements about sharing in "application
> subsystems" (type 3).
>
> Maybe you could give me a bit more detail on how you seen the
> increments working?
>
> Regards, Graham.
>
> On 25 February 2010 11:52, David Bosschaert <[email protected]> 
> wrote:
>> Right, the defaults would be different, but I guess there should be a
>> way to deviate from the defaults. Would it not make sense to use the
>> same syntax in use cases 1/3 to describe the deviation from the
>> defaults to what you are using in use case 2?
>>
>> What I'm thinking is: nothing specified -> defaults
>> something specified -> overrides defaults
>>
>> I don't fully understand why you would use different configuration
>> settings for this in options 1/3 just yet...
>>
>> Cheers,
>>
>> David
>>
>> On 25 February 2010 11:16, Graham Charters <[email protected]> wrote:
>>> Hi David,
>>>
>>> I think I see it a little differently.  If we try to common up the
>>> headers, then we end up with a matrix of rules for which headers are
>>> allowed on which type of 'subsystem'.  Take the example you gave
>>> below; for the three types of subsystem I described earlier, only the
>>> second type (Explicitly shared) would ever make statements about the
>>> services that are exported or imported.  Types 1 and 3 have different
>>> rules for defaulting what is shared.
>>>
>>> Regards, Graham.
>>>
>>> On 22 February 2010 10:24, David Bosschaert <[email protected]> 
>>> wrote:
>>>> I'm not entirely sure what you're thinking yet, could you maybe give
>>>> us an example?
>>>>
>>>> I would imagine that configuration entities with the same meaning
>>>> would be configured in the same way across the different types. So you
>>>> define the services that you're exporting using the same header/tag,
>>>> whether this is a subsystem or an application. Or do you see this
>>>> differently?
>>>>
>>>> Best regards,
>>>>
>>>> David
>>>>
>>>> On 19 February 2010 16:04, Graham Charters <[email protected]> 
>>>> wrote:
>>>>> I'd been thinking the different sets of manifest.  I think the way
>>>>> these types of subsystems will be used will be quite different and
>>>>> subsystem definitions will typically not morph from one type to
>>>>> another.  It therefore seems to make sense to emphasise the
>>>>> distinction.
>>>>>
>>>>> On 19 February 2010 15:01, Guillaume Nodet <[email protected]> wrote:
>>>>>> On Fri, Feb 19, 2010 at 15:00, Graham Charters 
>>>>>> <[email protected]> wrote:
>>>>>>> <snip>
>>>>>>> I think what we have so far is the basics of 3.  We should aim for
>>>>>>> consistency across all three, but I think the sharing policy defaults
>>>>>>> need to remain separate.  If we were to choose just one policy, then
>>>>>>> we will force the others into expressing a lot of unnecessary
>>>>>>> information.  We could broaden Application to cover all three, but I
>>>>>>> think that would be confusing.  Maybe there are other forms of
>>>>>>> subsystem for the different sharing policies, where each is
>>>>>>> specialized for the useful defaults.
>>>>>>
>>>>>> I agree with having different default policies.  What do you have in
>>>>>> mind as to identify those different use cases from a user point of
>>>>>> view ?  Are you thinking about completely different set of manifest
>>>>>> headers ? Or simply one which would contain the "kind" of
>>>>>> application/subsystem defined ?
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Cheers,
>>>>>> Guillaume Nodet
>>>>>> ------------------------
>>>>>> Blog: http://gnodet.blogspot.com/
>>>>>> ------------------------
>>>>>> Open Source SOA
>>>>>> http://fusesource.com
>>>>>>
>>>>>
>>>>
>>>
>>
>

Reply via email to