Re: Review of Row Level Security

On Wed, Dec 19, 2012 at 1:58 PM, Simon Riggs <simon(at)2ndquadrant(dot)com> wrote:
> If we don't enforce rules on INSERT the user has to specifically add a
> trigger, which makes things noticeably slower. There is more
> maintenance work for the average user, less performance and more
> mistakes to make.

Well, again, only if that's the behavior they want.

Also, it's also worth noting that, even if we assume that it is in
fact the behavior that users will want, the contention that it is
faster than a trigger is thus far unsubstantiated by any actual
benchmarks. It may indeed be faster ... but I don't know without
testing whether it's slightly faster or a whole lot faster. That
might be a good thing to find out, because if it is a whole lot
faster, that would certainly strengthen the case for including a mode
that works that way, whether or not we also provide other options.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company