Re: [Treebase-devel] Resolving the hibernate_sequence problem & another db instance

[Vladimir G. <vla...@du...>]

On Jan 7, 2010, at 11:08 AM, Rutger Vos wrote:

>> The plan, as Hilmar, you, and I decided at the beginning of the
>> conference call on Monday, is to re-direct postgres tables to use
>> hibernate_sequence, instead of their individual sequences, for new PK
>> IDs.  (The other solution we discussed would be to  change Java code
>> (or Hibernate XML mappings?) to use the tables' individual sequences
>> instead of hibernate_sequence.)
>
> Wait, I'm confused, is that the plan? I thought the plan was the other
> way around: we use the tables' individual sequences, which means we
> set the hibernate mode for generating IDs to "NATIVE" instead of
> "SEQUENCE". Am I the only one who understood this exactly the wrong
> way around?

That's what transpired from our discussion on Monday, before other  
people called in; we felt that prior email discussion did not reach a  
definite conclusion.

The outcome was influenced, as I recall, by Youjun's and Hilmar's  
observations that using a common sequence for all the tables is common  
practice in the Oracle world and a necessity in the MySQL world  
(keeping in mind a possibility of switching away from Postgres in the  
future), as well as the fact that this is how TB code works now and  
were not sure that nothing depended on this arrangement.

A volunteer to resume and lead the discussion to determine the more  
correct among the two (or more?) solutions?

My only reason for having held Youjun from proceeding one way or the  
other was the concern for messing up the only db instance we had.   
We'll soon have additional instances, so this will not be a roadblock.

--Vladimir