* Robert Haas (robertmhaas@gmail.com) wrote:
> Yes: let's punt this to 9.6. The decisions you're making here are way
> too significant to be making a couple of days before feature freeze,
> and this patch has changed massively since it was first submitted.
> There isn't time now for people who want to have an opinion on this to
> form an educated one.
Perhaps I'm missing something, but the patch has been simplified down to
the point where the only question seems to be "should we have default
roles or not?", which I had asked about two weeks ago and again last
week on a new thread. I feel like we're waiting for the silent majority
to chime in.
Put another way, I'm afraid that posting this next week, next month, or
next year is going to garner just as many responses as it's seen in the
past 2 weeks, while I continue to field questions on -bugs, -admin, and
IRC about "how do I set up Nagios with a non-superuser account?" and
similar issues. It's not a novel idea, certainly; Magnus suggested it
back in December on the thread, Tom made a similar comment that it might
make sense to have them later on and it's come up quite a few times
previously as it's something other RDBMS's have and we don't. Clearly,
others have read the proposal, at least (You and Alvaro on the other
thread, Heikki on this one).
It's my fault that I didn't follow up on their suggestions earlier and
instead spent a bunch of time fighting with pg_dump, but it doesn't seem
like there is a lot of disagreement about the idea. I'd offer to
simplify it down, but it seems like the obvious change in that direction
would be to reserve pg_ as a role prefix and not actually create any
default roles, but that doesn't gain us anything and makes a potential
headache for users without any feature to go with it.
Bruce's point is a better one, except that all of the changes have been
about reducing changes to core, down to an almost trivial level. I wish
it had been a smoother ride to get here from the original proposal six
months ago, but I've certainly got a better understanding of the level
of effort involved and changes required for the other approaches and
this certainly seems like the best and simplest.
Thanks!
Stephen