Hannu Krosing <hannu(at)krosing(dot)net> writes:
> Can't we keep a separate text .conf file specifically for the background
> processes which can't read system catalogs. It could contain only the
> GUCs these processes are interested in.
What's the value of that, compared to the existing proposal for
write-a-text-file-directly? It seems like useless complication.
If we could move *all* the GUCs into system catalogs, maybe it'd be
worth the trouble, but I think that's a fundamentally bad idea.
It will make it impossible to change settings when the system is down,
and thus for example impossible to fix configuration errors that
prevent the postmaster from starting. I think we should stick with
the principle that the text file is the primary authority, and that
means we don't need a system catalog entry for global settings.
A possibly instructive precedent is that we got rid of
pg_tablespace.spclocation after deciding it was counterproductive
to have a catalog entry that wasn't the authoritative state.
regards, tom lane
In response to
pgsql-hackers by date
|Next:||From: Atri Sharma||Date: 2012-11-16 19:42:26|
|Subject: Re: WIP patch for hint bit i/o mitigation|
|Previous:||From: Hannu Krosing||Date: 2012-11-16 19:06:14|
|Subject: Re: another idea for changing global configuration settings