I think we need to decide to do a full conversion or just leave it
for 1.0. If we are doing a full conversion, we need to come to
agreement on GBeanName and the query system.
-dain
On Aug 27, 2005, at 11:59 AM, Aaron Mulder wrote:
I agree that there's a lot involved here. I'd be OK with
providing the new and improved GBeanName implementation for M5 and
planning to do the total ObjectName->GBeanName conversion
afterward. I'd
also be OK with planning to do it all in M5 if everyone else is on
board
with that. I don't really like the remove/revert for M5 solution
-- just
because the feature is not complete and perfect does not mean we
shouldn't
make incremental progress (and believe me, I feel like I've been on
the
pointy end of that one before).
Aaron
On Sat, 27 Aug 2005, Dain Sundstrom wrote:
How about a must have to implement GBeanName according to the
previous notes on the mailing list?
Does this include modifying all code to use GBeanName instead of
object name? If not, I think we should simply remove GBeanName
instead because it makes the kernel confusing. The Kernel interface
has methods that take object names, and if a subset of ObjectNames
are invalid for the kernel this interface is misleading. Also the
only use of GBeanName in the kernel is within the registry code.
This means that the rest of the framework assumes ObjectNames, and
this change will make that code confusing. Finally, we have not
addressed ObjectName queries, which are a required component of the
framework and are used through the code base. This should be an all
or nothing change.
-dain
