On 12 March 2017 at 06:51, Joe Conway <mail@joeconway.com> wrote:
> My opinion is that the user visible aspects of this should be deprecated
> and correct syntax provided. But perhaps that is overkill.
FWIW, in my experience, pretty much nobody understands the pretty
tangled behaviour of "WITH [ENCRYPTED] PASSWORD", you have to
understand the fact table of:
* ENCRYPTED, UNENCRYPTED or neither set
* password_encryption GUC on or off
* password begins / doesn't begin with fixed string 'md5'
to fully know what will happen.
Then of course, you have to understand how all this interacts with
pg_hba.conf's 'password' and 'md5' options.
It's a right mess. Since our catalogs don't keep track of the hash
separately to the password text and use prefixes instead, and since we
need compatibility for dumps, it's hard to do a great deal about
though.
I'm not convinced that a keyword change will do much good, the whole
thing really needs a reassessment to make sure that it's clearer to
users/admins and has fewer moving parts.
So I'm in favour of fixing the docs but I'm not keen on changing the
SQL syntax in a way that just kind of papers over part of the
problems.
-- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services