Re: reloptions with a "namespace"

Euler Taveira de Oliveira wrote:
> Alvaro Herrera escreveu:
> >> IIRC, my last patch includes a partial validation code for RESET cases. For
> >> example, the last SQL will not be atomic (invalid reloption silently ignored).
> >> So, why not apply the namespace validation code to RESET case too? Patch is
> >> attached too.
> > 
> > No, we must not validate the options passed to RESET, because we want to
> > be able to reset even options that we do not currently think that are
> > valid.  Consider that we might be trying to clean up after options set
> > by a previous version of a module.
> > 
> Ah, idea withdrawn. But we should at least document this behavior.

Well, it is documented -- see amoptions here
http://www.postgresql.org/docs/8.3/static/index-functions.html

The problem with this is that documentation for reloptions is scattered
all over the place.  I think we should have a separate section somewhere
on which they are discussed at length.

-- 
Alvaro Herrera                                http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.