On 3/29/2012 3:32 PM, Katherine Marsden wrote:
On 3/29/2012 12:44 PM, Mike Matrigali wrote:
We do continue to keep adding to problem here. This is yet another
catalog that is added to db, and everyone has to pay for it even if they
don't use the sequence feature. I know "1" more does not sound like
a problem, but I don't think it is the right direction for derby.
I know that there are applications out there that use 1000's of
databases. We have definitely gotten complaints from users in the
past just for having 1 background thread per database. Every
resource we add
on a per database level can get multiplied in some applications, and it
is sad to add these overheads when the application is not even using the
feature. I have the same problem with the new authorization work. I
understand that the system has current problems in this area (roles),
but we should not keep digging bigger holes.
+1
I think we have to think of stability, supportability, compatibility,
and scalability first especially in areas that affect existing
functionality that is heavily used. The native authentication work is
at least isolated if not used but the increasing complexity and issues
around sequences are interjected into identity columns. That makes them
of much greater concern because of the wide use and dependence under
stress conditions in the field.
I put a comment in DERBY-5671 regarding this concern[1]. Disconnecting
identity columns from sequences is really necessary at this point,
because it appears both the existing problems with sequences and
whatever path sequences go down to correct them are too dangerous for
an identity column hot swap.
Thanks
Kathey
[1]https://issues.apache.org/jira/browse/DERBY-5671
