Hi, On 2018-12-09 15:42:57 -0500, Stephen Frost wrote: > * Tom Lane (t...@sss.pgh.pa.us) wrote: > > John Naylor <jcnay...@gmail.com> writes: > > > Commit 96cdeae07 added toast tables to most catalogs. One disadvantage > > > is that the toast declarations require hard-coded oids, even though > > > only shared catalogs actually need stable oids. Now that we can assign > > > oids on the fly, it makes sense to do so for toast tables as well, as > > > in the attached. > > > > I'm a bit dubious that this is a good idea. It's handy, at least for > > forensic situations, that the system catalogs have stable OIDs.
Hm, but won't they have that for major versions anyway? We ought not to change the .bki generation in a way that results in differing oids after a release, no? > I tend to agree... What's the advantage of assigning them on the fly? No idea if that's John's reasoning, but I do like not having to do yet another manual step that you need to remember/figure out when adding a new catalog relation. Greetings, Andres Freund
