Hi Mike,
On 2/18/11 11:06 AM, Danny van der Rijn wrote:
4D5EC330.60708@tibco.com" type="cite"><vdR2>
I agree that
mechanically there is little difference between an unsatisfied
reference and an unsatisfied GDC. Semantically, however,
there is quite a difference. In order to mechanically satisfy
the unsatisfied GDC (albeit minimally) all that is required is
a name, and the name is known exactly by the unsatisfied use
of the GDC. Therefore the runtime already has enough
information to head off the error. Not so in the case of an
unsatisfied reference.
</vdR2>
OF29957524.44368224-ON80257838.003AB000-80257838.003B4D32@uk.ibm.com"
type="cite">
<mje>
Thank you for
putting it so clearly.
My concerns with
auto-creation surround this possibility of deployment taking
place as a series of steps, rather than as a "single big
bang" - and with the question of what happens at the
completion of each step.
I am concerned
with the case where the Domain level channel is there
*somewhere* in the deployed contributions but where it has
not yet been deployed. Allowing auto-creation in these
circumstances is problematic, if auto-creation occurs before
the deployment of the Domain level channel.
</mje>
At least for this point, this is one of the reasons that I filed
http://osoa.org/jira/browse/ASSEMBLY-257
If this matters to you, perhaps we should address the question of
how they're deployed before worrying about whether they can be
auto-deployed?
-Eric.
|