Re: Switching roles as an replacement of connection pooling tools

Поиск
Список
Период
Сортировка
От Francisco Olarte
Тема Re: Switching roles as an replacement of connection pooling tools
Дата
Msg-id CA+bJJbxqBQbG0RpQq3ATrufZ0d7AYWn+RZtH5y87tdS-YUH1BQ@mail.gmail.com
обсуждение исходный текст
Ответ на Re: Switching roles as an replacement of connection pooling tools  (Melvin Davidson <melvin6925@gmail.com>)
Список pgsql-general
Hi Melvin:

On Tue, May 31, 2016 at 3:55 PM, Melvin Davidson <melvin6925@gmail.com> wrote:
> On Tue, May 31, 2016 at 3:45 AM, CN <cnliou9@fastmail.fm> wrote:
>> SET SESSION AUTHORIZATION user2 PASSWORD p2;

> Your points make no sense. You can accomplish the same with:
> GRANT ROLE user2 TO user1;

I'm not discussing wether it makes sense, but you do not acomplish the
same. In his case you cannot swict to the other role unless you know
the password for it, in yours you can. Also I suspect he wants it to
work like a login, i.e., if you have N roles and you add another one
he wants it to be like a new login user, and apps/people could have a
set of X role+password combos different for each one. I think it's a
bizarre thing, but not the same as granting some roles to other ( of
course if you have M people and N schemas ( in his example )you can
have M login roles and grant combos of N 'schema roles' to them to
achieve this, but if N is, say, a hundred, and you have a huge M, like
ten thousand, with a different combo for each one, his solution may
make sense ( I do not think such a bizarre case justifies the bug-risk
of including the feature, but it can make sense ) )

Francisco Olarte.


В списке pgsql-general по дате отправления:

Предыдущее
От: Thalis Kalfigkopoulos
Дата:
Сообщение: Re: Drop/Re-Creating database extremely slow + doesn't lose data
Следующее
От: Oleg Bartunov
Дата:
Сообщение: Re: Slides for PGCon2016; "FTS is dead ? Long live FTS !"