[hibernate-dev] HHH-7018 and one more change to how we serialize EntityManagerFactory instances...

[Steve Ebersole]

Ah, nice.  I did not even think of reusing SESSION_FACTORY_NAME for the 
EMF name, but that should work well.


On Wed 08 Feb 2012 01:26:30 PM CST, Scott Marlow wrote:
> I'll set the EntityManagerFactory name to:
>
> (1) explicit AvailableSettings.SESSION_FACTORY_NAME if specified
> (2) use application specified pu-name if pu-name is specified (not 
> always the case, especially in test environments)
> (3) uuid
>
>
> On 02/08/2012 11:08 AM, Steve Ebersole wrote:
>> If I understand correctly you are thinking there is a case where the
>> "PU name" would not be unique for a given app? That is the only time
>> I can see this being a concern. If it is possible that the same app
>> can have different PU names on different nodes in the cluster then
>> simply using PU name as EMF name wont work there. But pretty sure I
>> said that this will need to allow explicitly setting of the EMF name.
>>
>> But regardless UUID will never work in a cluster. Thats been covered
>> a few times here. UUID is merely a fallback which assumes ser and
>> deser happen in the same VM.
>>
>> On Wed, Feb 8, 2012 at 9:17 AM, Scott Marlow<smarlow at redhat.com> wrote:
>>> It wouldn't be difficult to switch back to using UUIDs in the EMF. 
>>> Neither
>>> way is perfect (see discussion on HHH-6897).
>>>
>>>
>>>
>>> On 02/08/2012 08:28 AM, Steve Ebersole wrote:
>>>>
>>>> On 02/08/2012 04:23 AM, Christian Bauer wrote:
>>>>>
>>>>> (For whatever reason, I still don't understand why EMF clustering 
>>>>> would
>>>>> be different than SF clustering.)
>>>>
>>>>
>>>> This is exactly the point I am missing here as well.
>>>>
>>>
>

-- 
steve at hibernate.org
http://hibernate.org