On November 8, 2004 09:55 am, Darcy Buskermolen wrote:
> Hello,
>
> In order to apply some more improvements to the 1.1 branch of slony making
> it more pg_dump friendly. I'm wondering if there is a compelling reason to
> continue to have sl_table.tab_reloid and sl_sequence.seq_reloid reference
> pg_class.oid, instead of storing the fully qualified name ?

Ok I've done some tests, and beat some code around and have pretty much 
discoverd that there is no "nice" way to convert to name usage and have ALTER 
TABLE foo RENAME bar work and not break replication under some cases.  So at 
this moment I'm inclined to extend sl_tab and sl_seq to include the relname 
and namespace (FQN). Add a new slonik command that can be used to update the 
sl_tab/seq reloids based on the FQN (and the reverse), and also run this 
after any execute script.  Having said this, I have an option of storing the 
FQN in 1 field, in the form of  "namespace"."relname", or we can break it out 
into 2 cols, one for the namespace, and one for the relname.

This is your chance to voice your opinions on this.

-- 
Darcy Buskermolen
Wavefire Technologies Corp.
ph: 250.717.0200
fx:  250.763.1759
http://www.wavefire.com