On 6/16/2006 5:38 PM, Christopher Browne wrote:
> Jan Wieck wrote:
>> On 6/12/2006 5:24 PM, Christopher Browne wrote:
>>> Jim C. Nasby wrote:
>>>> Moving to slony1-general...
>>>>
>>>> Hmm... is sl_trigger intentionally created WITH OIDs?
>>>>   
>>> None of the Slony-I internal tables are defined with reference to "WITH
>>> OIDS" or "WITHOUT OIDS".
>>>
>>> In effect, they will take on the local behaviour of the version of
>>> PostgreSQL on which they are defined.
>>>
>>> I don't think sl_trigger will be particularly interesting; it is a
>>> config table that could chew up a few OIDs, but is not frequently
>>> updated, so it won't chew up many.
>>>
>>> In contrast, sl_event, sl_log_1, sl_seqlog, sl_log_2, could all chew up
>>> plenty of OIDs if they are, by default, created with OIDs.
>>>
>>> Presumably one could alter these tables thus:
>>>  
>>>    alter table _myschema.sl_log_1 set without oids;
>>>
>>> I think I'd want to test this before doing it, and consider taking a
>>> brief application outage to implement it, assuming it worked fine in
>>> test...
>>
>> Slony itself doesn't depend on OID's, so it'd be safe to actually
>> create all Slony tables explicitly without oid's.
>>
> Interesting.
> 
> Then it could be a good change to deploy in general, to specify all
> tables in the Slony-I schema as "WITHOUT OIDS".
> 
> I'm not sure it's wise to add it just as I'm hoping to start proposing
> release candidates for 1.2.0; do you think it would be a good idea to
> add this now??

I don't think 1.2.1 would be a better time.


Jan

-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #