Thread: DROP OWNED BY
I'm working on the promised DROP OWNED patch, and just noticed that it needs modifications to the Grant/Revoke machinery that are too invasive to consider for 8.1; the problem is that ExecuteGrantStmt takes textual names for objects and grantees, and what I have from pg_shdepend are Oids. Dropping the objects themselves is easy -- the problem is dropping grants to other objects. So, instead of doing that, I think it's better to convert the Oids to names in the shared-dependency code. This should be rewritten after the next development period starts, for efficiency and cleanliness. Is anyone opposed to this idea? Further, is anyone opposed to the addition as a whole? I'm already a month and a half late with this part of the shdep patch, so if people think it's better to ship this release without it, so be it. (Of course I don't like that.) -- Alvaro Herrera (<alvherre[a]alvh.no-ip.org>) "Cuando no hay humildad las personas se degradan" (A. Christie)
Definitely should be released without it ... it is an extension not submitted for inclusion prior to the feature freeze ... On Fri, 12 Aug 2005, Alvaro Herrera wrote: > I'm working on the promised DROP OWNED patch, and just noticed that it > needs modifications to the Grant/Revoke machinery that are too invasive > to consider for 8.1; the problem is that ExecuteGrantStmt takes textual > names for objects and grantees, and what I have from pg_shdepend are > Oids. Dropping the objects themselves is easy -- the problem is > dropping grants to other objects. > > So, instead of doing that, I think it's better to convert the Oids to > names in the shared-dependency code. This should be rewritten after the > next development period starts, for efficiency and cleanliness. > > Is anyone opposed to this idea? Further, is anyone opposed to the > addition as a whole? I'm already a month and a half late with this > part of the shdep patch, so if people think it's better to ship this > release without it, so be it. (Of course I don't like that.) > > -- > Alvaro Herrera (<alvherre[a]alvh.no-ip.org>) > "Cuando no hay humildad las personas se degradan" (A. Christie) > > ---------------------------(end of broadcast)--------------------------- > TIP 5: don't forget to increase your free space map settings > > > ---- Marc G. Fournier Hub.Org Networking Services (http://www.hub.org) Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664
Alvaro Herrera <alvherre@alvh.no-ip.org> writes: > I'm working on the promised DROP OWNED patch, and just noticed that it > needs modifications to the Grant/Revoke machinery that are too invasive > to consider for 8.1; the problem is that ExecuteGrantStmt takes textual > names for objects and grantees, and what I have from pg_shdepend are > Oids. So refactor the GrantStmt code into a layer that translates to OIDs and another layer that does the work. The other way doesn't sound good to me; it is prone to race conditions against someone doing a RENAME. > Is anyone opposed to this idea? Further, is anyone opposed to the > addition as a whole? I'm already a month and a half late with this > part of the shdep patch, so if people think it's better to ship this > release without it, so be it. (Of course I don't like that.) I think we're rapidly approaching the point where we have to say "sorry, that is too late for this release". DROP OWNED is really a new feature, not an essential component of the ROLES patch, and we've already stretched the no-new-features-after-freeze rule to the breaking point. I'll take the full blame for that --- I should probably not have forced in the ROLES patch so close to feature freeze when it was obviously not finished --- but at some point we've got to say no. regards, tom lane
On Sat, Aug 13, 2005 at 10:09:35AM -0400, Tom Lane wrote: > Alvaro Herrera <alvherre@alvh.no-ip.org> writes: > > I'm working on the promised DROP OWNED patch, and just noticed that it > > needs modifications to the Grant/Revoke machinery that are too invasive > > to consider for 8.1; the problem is that ExecuteGrantStmt takes textual > > names for objects and grantees, and what I have from pg_shdepend are > > Oids. > > So refactor the GrantStmt code into a layer that translates to OIDs and > another layer that does the work. The other way doesn't sound good to > me; it is prone to race conditions against someone doing a RENAME. Roger. I'll produce a patch for 8.2. -- Alvaro Herrera (<alvherre[a]alvh.no-ip.org>) "Es filósofo el que disfruta con los enigmas" (G. Coli)