On Tue, Jan 27, 2009 at 8:49 AM, Bernd Helmle <mailings@oopsware.de> wrote:
> --On Dienstag, Januar 27, 2009 14:04:05 +0200 Peter Eisentraut
> <peter_e@gmx.net> wrote:
>
>>> a view should be updatable by default if the query expression is
>>> updatable... what we need is something to make a view READ ONLY even
>>> if it should be updatable by spec...
>>
>> A view is read-only if you don't grant any write permissions on it.
>
> What i'm seeing here is a very divergent understanding what a "read-only"
> view is:
>
> old-school PostgreSQL-Users would expect a "read-only" view to have no
> "write action" installed. If we want to follow the standard closely, they
> need to be installed automatically, changing this behavior, hence the wish
> to have a syntax to restore the old behavior (e.g. for pg_dump). I'm unsure
> what the correct approach looks like, but it seems we need a compromise.
Do we REALLY think there are people out there who are writing INSERT
or UPDATE actions on views on which they haven't installed rules and
counting on the fact that those operations fail for correctness?
Personally, I usually write my code so it inserts into something that
is, uh... insertable.
...Robert