Users/Groups -> Roles

Greetings,

Attached please find files and patches associated with moving from the
User/Group system currently in place to Roles, as discussed
previously. The files are:

pg_authid.h
New system table, contains role information
To be placed in src/include/catalog, replacing pg_shadow.h

pg_auth_members.h
New system table, contains role/membership information
To be placed in src/include/catalog, replacing pg_group.h

role_2005062701.ctx.patch.bz2
Main patch, generated via cvs -z3 diff -c | bzip2
(gzip didn't quite get it under the 100K mark for the list)

role_milestones
List of milestones associated with my work on Roles support
'*' indicates that the milestone has been met/completed
'?' indicates uncertainty about if something should be done
No marker indicates an item which needs to be done
Note: Documentation needs to be updated

Again, my apologies for not getting this in sooner, it's been a little
hectic around here of late. I'm anxious to get feedback, comments,
corrections, etc.

Thanks,

Stephen

Greetings,

(Sent this earlier, but afraid it may have gotten caught by the
too-big bug, so I'm reposting without the files attached, they can
all be found at: http://kenobi.snowman.net/~sfrost/pg_role/ ; there
are also gzip and uncompressed versions of the unified / context
diffs there for those who don't care for bzip2)

Attached please find files and patches associated with moving from the
User/Group system currently in place to Roles, as discussed
previously. The files are:

pg_authid.h
New system table, contains role information
To be placed in src/include/catalog, replacing pg_shadow.h

pg_auth_members.h
New system table, contains role/membership information
To be placed in src/include/catalog, replacing pg_group.h

role_2005062701.ctx.patch.bz2
Main patch, generated via cvs -z3 diff -c | bzip2
(gzip didn't quite get it under the 100K mark for the list)

role_milestones
List of milestones associated with my work on Roles support
'*' indicates that the milestone has been met/completed
'?' indicates uncertainty about if something should be done
No marker indicates an item which needs to be done
Note: Documentation needs to be updated

Again, my apologies for not getting this in sooner, it's been a little
hectic around here of late. I'm anxious to get feedback, comments,
corrections, etc.

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> Attached please find files and patches associated with moving from the
> User/Group system currently in place to Roles, as discussed
> previously.

I have cleaned this up a bit and committed it. I normally wouldn't
commit an incomplete patch, but this change is blocking Alvaro's work
on dependencies for shared objects, so I felt it was best to get the
catalog changes in now. That will let Alvaro work on dependencies
while I sort out the unfinished bits of roles, which I intend to do
over the next day or so.

Many thanks for your work on this!

regards, tom lane

Dear Stephen,

> Attached please find files and patches associated with moving from the
> User/Group system currently in place to Roles, as discussed
> previously. The files are:
>
> pg_authid.h
> New system table, contains role information
> To be placed in src/include/catalog, replacing pg_shadow.h
>
> pg_auth_members.h
> New system table, contains role/membership information
> To be placed in src/include/catalog, replacing pg_group.h

I've looked very quickly at the patch. ISTM that the proposed patch is a
reworking of the user/group stuff, which are both unified for a new "role"
concept where a user is a kind of role and a role can be a member of
another role. Well, why not.

Some added files seems not to be provided in the patch :

sh> bzgrep pg_authid.h ./role_2005062701.ctx.patch.bz2
? src/include/catalog/pg_authid.h
! pg_namespace.h pg_conversion.h pg_database.h pg_authid.h pg_auth_members.h \
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
+ #include "catalog/pg_authid.h"
+ #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"
! #include "catalog/pg_authid.h"

Or maybe I missed something, but I could not find the pg_authid file?
the '?' line in the diff seems to suggest an unexpected file.

Anyway, from what I can see in the patch it seems that the roles are per
cluster, and not per catalog. So this is not so conceptually different
from user/group as already provided in pg.

What would have been much more interesting for me would be a per catalog
role, so that rights could be administrated locally in each database. I'm
not sure how to provide such a feature, AFAICS the current version does
not give me new abilities wrt right management.

Have a nice day,

--
Fabien.

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > Attached please find files and patches associated with moving from the
> > User/Group system currently in place to Roles, as discussed
> > previously.
>
> I have cleaned this up a bit and committed it. I normally wouldn't
> commit an incomplete patch, but this change is blocking Alvaro's work
> on dependencies for shared objects, so I felt it was best to get the
> catalog changes in now. That will let Alvaro work on dependencies
> while I sort out the unfinished bits of roles, which I intend to do
> over the next day or so.

Great, glad to hear it. I hope you got a chance to look over the open
items in the 'milestones' file. I'd really like to see the grammar be
fixed to match SQL spec for GRANT ROLE/REVOKE ROLE. I think an approach
to take there might be to try and get GrantRoleStmt and GrantStmt to use
the same productions at the end of the line if possible or something
along those lines.

Also, I've been looking through the diff between my tree and what you
committed to CVS and had a couple comments (just my 2c: I think it would
have been alot easier using SVN to see exaclty what was different from
my patch vs. other changes since my last CVS up):

First, sorry about the gratuitous name changes, it helped me find
every place I needed to look at the code and think about if it needed
to be changed in some way (ie: Int32GetDatum -> ObjectIdGetDatum,
etc). I had planned on changing some of them back to minimize the
patch but kind of ran out of time.

Second, looks like I missed fixing an owner check in pg_proc.c
Current CVS has, line 269:
if (GetUserId() != oldproc->proowner && !superuser())
Which is not a sufficient owner check. This should by fixed by doing
a proper pg_proc_ownercheck, ie:
if (!pg_proc_ownercheck(HeapTupleGetOid(oldtup), GetUserId()))

Third, I feel it's incorrect to only allow superuser() to change
ownership of objects under a role-based system. Users must be able to
create objects owned by a role they're in (as opposed to owned only
by themselves). Without this there is no way for a given role to
allow other roles to perform owner-level actions on objects which they
create. The point of adding roles was to allow owner-level actions on
objects to more than a single user or the superuser. Requiring the
superuser to get involved with every table creation defeats much of
the point.

This should really be possible either by explicitly changing the
ownership of an object using ALTER ... OWNER, or by a SET ROLE
followed by CREATE TABLE, etc. SET ROLE is defined by the SQL
specification, though we don't support it specifically yet (shouldn't
be too difficult to add now though). Certainly if we accept that
SET ROLE should be supported and that objects then created should be
owned by the role set in SET ROLE we should be willing to support
non-superusers doing ALTER ... OWNER given that they could effectively
do the same thing via SET ROLE (though with much more difficulty,
which has no appreciable gain).

Fourth, not that I use it, but, it looks like my changes to
src/interfaces/ecpg/preproc/preproc.y were lost. Not sure if that was
intentional or not (I wouldn't think so... I do wish ecpg could just
be the differences necessary for ecpg and be based off the main parser
somehow, but that'd be a rather large change). Oh, and in that same
boat, src/tools/pgindent/pgindent also appears to not have gotten the
changes that I made.

> Many thanks for your work on this!

Happy to have helped though frustrated that you seem to have removed
the part that I was originally looking for. I don't feel that's
justification for having it (I feel I've addressed that above) but it
certainly would have been nice to be aware of that earlier and perhaps
to have discussed the issues around it a bit more before being so
close to the feature freeze (I know, alot my fault, but there it is).

Thanks,

Stephen

Hi Fabien,

* Fabien COELHO (fabien(dot)coelho(at)ensmp(dot)fr) wrote:
> I've looked very quickly at the patch. ISTM that the proposed patch is a
> reworking of the user/group stuff, which are both unified for a new "role"
> concept where a user is a kind of role and a role can be a member of
> another role. Well, why not.

Right, it's a beginning to proper 'Role' support as defined by the SQL
specification.

> Some added files seems not to be provided in the patch :

pg_authid.h and pg_auth_members.h were attached to the email. They're
also available at http://kenobi.snowman.net/~sfrost/pg_role/ ; but the
patch has already been applied by Tom to CVS HEAD (well, with lots of
modifications and whatnot), so you probably should just take a look at
that.

> Anyway, from what I can see in the patch it seems that the roles are per
> cluster, and not per catalog. So this is not so conceptually different
> from user/group as already provided in pg.

It's conceptually different from users/groups in that it's roles, which
aren't the same thing. You're right, it's still per-cluster though.

> What would have been much more interesting for me would be a per catalog
> role, so that rights could be administrated locally in each database. I'm
> not sure how to provide such a feature, AFAICS the current version does
> not give me new abilities wrt right management.

I understand your concerns here and while I agree with the basic idea
that per-catalog role sets would be nice it wasn't what I had set out to
do with this patch. Perhaps what you're asking for will be added later
on. Some things this patch does do though are:

Allow role ownership. This role can also have members, and doesn't
necessairly have to be allowed to log in. Members of a role which owns
an object have owner-level rights on that object (so, fe: roles user1,
user2 and group1 where user1 and user2 are members of group1, a table
owned by group1 can be vacuumed, have columns added/removed, have
indexes create on it, etc, by user1 or user2).

Allow granting roles to other roles based on the 'with admin option'.
This means you don't have to be a superuser to add a member to a role
which you have the 'admin option' on.

There's other things (startup may be a bit faster since the pg_auth file
is sorted by the backend instead of during each startup, etc) but the
above were the types of things that I was looking to do mainly.

I'd like to see it possible to distinguish between 'superuser' and
'createrole' permissions, but I didn't get to that point with the roles
support (it's really a seperate issue anyway).

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> Also, I've been looking through the diff between my tree and what you
> committed to CVS and had a couple comments

> First, sorry about the gratuitous name changes, it helped me find
> every place I needed to look at the code and think about if it needed
> to be changed in some way (ie: Int32GetDatum -> ObjectIdGetDatum,
> etc). I had planned on changing some of them back to minimize the
> patch but kind of ran out of time.

No problem, I figured that was why you'd done it, but changing them back
helped me to understand the patch also ;-)

> Second, looks like I missed fixing an owner check in pg_proc.c

Got it. I was wondering if there were more --- might be worth checking
all the superuser() calls.

> Third, I feel it's incorrect to only allow superuser() to change
> ownership of objects under a role-based system.

I took that out because it struck me as a likely security hole; we don't
allow non-superuser users to give away objects now, and we shouldn't
allow non-superuser roles to do so either. Moreover the tests you had
were inconsistent (not same test everyplace).

> Users must be able to
> create objects owned by a role they're in (as opposed to owned only
> by themselves).

This is what SET SESSION AUTHORIZATION/SET ROLE is for, no? You set the
auth to a role you are allowed to be in, then create the object. I do
notice that we don't have this yet, but it's surely a required piece of
the puzzle.

> Fourth, not that I use it, but, it looks like my changes to
> src/interfaces/ecpg/preproc/preproc.y were lost. Not sure if that was
> intentional or not

Yeah, it was. I leave it to Michael Meskes to sync ecpg with the main
parser; on the occasions where I've tried to do it for him, things
didn't work out well.

> I do wish ecpg could just
> be the differences necessary for ecpg and be based off the main parser
> somehow,

Me too, but I haven't seen a way yet.

> src/tools/pgindent/pgindent also appears to not have gotten the
> changes that I made.

That's an automatically generated list; there's no need to edit it.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > Second, looks like I missed fixing an owner check in pg_proc.c
>
> Got it. I was wondering if there were more --- might be worth checking
> all the superuser() calls.

Yeah, let's come up with a decision about what exactly we should allow
and then perhaps I can go through all of the superuser() calls and see
what needs fixing up.

> > Third, I feel it's incorrect to only allow superuser() to change
> > ownership of objects under a role-based system.
>
> I took that out because it struck me as a likely security hole; we don't
> allow non-superuser users to give away objects now, and we shouldn't
> allow non-superuser roles to do so either. Moreover the tests you had
> were inconsistent (not same test everyplace).

Sorry about them being inconsistent, I didn't intend for them to be.
I went through a couple of iterations of them trying to do the check the
'right' way. Thinking back on it, even the checks I ended up with were
wrong (in the superuser case), though I think they were closer.
Basically my thought was to allow the same thing you could do w/ SET
ROLE, etc:

If you are the owner of the object to be changed (following the normal
owner checking rules) AND would still be considered the owner of the
object *after* the change, then you can change the ownership.

The code I had for this was:

if (!pg_class_ownercheck(tuple,GetUserId()) ||
!is_role_member(newowner,GetUserId()))

That needs a check for superuser though because while the test will pass
on the 'pg_class_ownercheck' side, it won't on the 'is_role_member' side
(currently anyway, I suppose a superuser could be considered to be in
any role, so we could change is_role_member to always return true for
superusers, that'd probably make pg_group look ugly though, either way):

if (!superuser() && !(pg_class_ownercheck(tupe,GetUserId()) &&
is_role_member(newowner,GetUserId())))

I think that's the correct check and can be done the same way for pretty
much all of the objects. Were there other security concerns you had?
I'd be happy to look through the superuser() checks in commands/ and
develop a patch following what I described above, as well as looking for
other cases where we should be using the *_ownercheck() functions.

One place I recall seeing one and not being sure if it should be a new
*_ownercheck() function or not was in the 2PC patch- twophase.c, line
380:

if (user != gxact->owner && !superuser_arg(user))

Wasn't sure if that made sense to have *_ownercheck, or, even if we
added one for it, if it made sense to check is_member_of_role() for
prepared transactions. I don't think SQL has anything to say about it,
anyone know what other DBs do here?

> > Users must be able to
> > create objects owned by a role they're in (as opposed to owned only
> > by themselves).
>
> This is what SET SESSION AUTHORIZATION/SET ROLE is for, no? You set the
> auth to a role you are allowed to be in, then create the object. I do
> notice that we don't have this yet, but it's surely a required piece of
> the puzzle.

(Technically I think SET SESSION AUTHORIZATION is different from SET
ROLE, but anyway)

Right, that's another way to do it (as I mentioned), and that lets you
do ownership changes, but they're much more painful:

CONNECT AS joe;
CREATE TABLE abc as SELECT name,avg(a),sum(b) FROM reallybigtable;
-- Whoops, I meant for abc to be owned by role C so sally can add her
-- column to it later, or vacuum/analyze it, whatever
GRANT SELECT ON abc TO C; -- Might not be necessary
ALTER TABLE abc RENAME TO abc_temp;
SET ROLE C;
CREATE TABLE abc AS SELECT * FROM abc_temp; -- Could be big :(
SET ROLE NONE; -- Might be just 'SET ROLE;'? Gotta check the spec
DROP TABLE abc_temp;

I don't really see the point in making users go through all of these
hoops to do an ownership change. In the end, it's the same result near
as I can tell...

> Yeah, it was. I leave it to Michael Meskes to sync ecpg with the main
> parser; on the occasions where I've tried to do it for him, things
> didn't work out well.

Ah, ok.

> > src/tools/pgindent/pgindent also appears to not have gotten the
> > changes that I made.
>
> That's an automatically generated list; there's no need to edit it.

Hah, silly me.

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> The code I had for this was:

> if (!pg_class_ownercheck(tuple,GetUserId()) ||
> !is_role_member(newowner,GetUserId()))

> That needs a check for superuser though because while the test will pass
> on the 'pg_class_ownercheck' side, it won't on the 'is_role_member' side

Um, right, that was another problem I had with it --- at one point the
regression tests were failing because the superuser wasn't allowed to
reassign object ownership ...

I'm still fairly concerned about the security implications of letting
ordinary users reassign object ownership. The fact that SET ROLE would
let you *create* an object with ownership X is a long way away from
saying that you should be allowed to change an *existing* object to have
ownership X. This is particularly so if you are a member of a couple of
different roles with different memberships: you will be able to cause
objects to become effectively owned by certain other people, or make
them stop being effectively owned by those people. I don't have a clear
trouble case in mind at the moment, but this sure sounds like the stuff
of routine security-hole reports. (Altering the ownership of a SECURITY
DEFINER function, in particular, sounds like a great path for a cracker
to pursue.)

> One place I recall seeing one and not being sure if it should be a new
> *_ownercheck() function or not was in the 2PC patch- twophase.c, line
> 380:

This one I think we can leave...

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > That needs a check for superuser though because while the test will pass
> > on the 'pg_class_ownercheck' side, it won't on the 'is_role_member' side
>
> Um, right, that was another problem I had with it --- at one point the
> regression tests were failing because the superuser wasn't allowed to
> reassign object ownership ...

Yeah, sorry about that.

> I'm still fairly concerned about the security implications of letting
> ordinary users reassign object ownership. The fact that SET ROLE would
> let you *create* an object with ownership X is a long way away from
> saying that you should be allowed to change an *existing* object to have
> ownership X. This is particularly so if you are a member of a couple of
> different roles with different memberships: you will be able to cause
> objects to become effectively owned by certain other people, or make
> them stop being effectively owned by those people. I don't have a clear
> trouble case in mind at the moment, but this sure sounds like the stuff
> of routine security-hole reports. (Altering the ownership of a SECURITY
> DEFINER function, in particular, sounds like a great path for a cracker
> to pursue.)

SET ROLE also lets you *drop* an object owned by that role. Or alter
it. Or CREATE OR REPLACE FUNCTION ...

I can understand your concern. The specific use case I'm thinking about
is where a user creates an object, does some work on it, and then wants
to change its ownership to be owned by a role which that user is in. I
find myself doing that a fair bit (as superuser atm). One thing I don't
like about limiting it to that is that you then can't go back without
the whole drop/create business or getting an admin.

This also isn't stuff that couldn't be done through other means, even in
the SECURITY DEFINER function case, you just need to drop, set role,
create. Having a role with members be able to own objects isn't meant
to replace the privileges system and I don't expect people to try to use
it to.

I can perhaps see a special case for SECURITY DEFINER functions but if
we're going to special case them I'd think we'd need to make them only
be creatable/modifiable at all by superusers or add another flag to the
role to allow that.

Thanks,

Stephen

On Tue, Jun 28, 2005 at 14:45:06 -0400,
Stephen Frost <sfrost(at)snowman(dot)net> wrote:
>
> If you are the owner of the object to be changed (following the normal
> owner checking rules) AND would still be considered the owner of the
> object *after* the change, then you can change the ownership.

That still isn't a good idea, because the new owner may not have had
access to create the object you just gave to them. Or you may not have
had access to drop the object you just gave away. That is going to
be a security hole.

On Tue, Jun 28, 2005 at 14:52:07 -0500,
Bruno Wolff III <bruno(at)wolff(dot)to> wrote:
> On Tue, Jun 28, 2005 at 14:45:06 -0400,
> Stephen Frost <sfrost(at)snowman(dot)net> wrote:
> >
> > If you are the owner of the object to be changed (following the normal
> > owner checking rules) AND would still be considered the owner of the
> > object *after* the change, then you can change the ownership.
>
> That still isn't a good idea, because the new owner may not have had
> access to create the object you just gave to them. Or you may not have
> had access to drop the object you just gave away. That is going to
> be a security hole.

Thinking about it some more, drops wouldn't be an issue since the owner
can always drop objects.

Creating objects in particular schemas or databases is not something that
all roles may be able to do.

* Bruno Wolff III (bruno(at)wolff(dot)to) wrote:
> On Tue, Jun 28, 2005 at 14:45:06 -0400,
> Stephen Frost <sfrost(at)snowman(dot)net> wrote:
> >
> > If you are the owner of the object to be changed (following the normal
> > owner checking rules) AND would still be considered the owner of the
> > object *after* the change, then you can change the ownership.
>
> That still isn't a good idea, because the new owner may not have had
> access to create the object you just gave to them. Or you may not have
> had access to drop the object you just gave away. That is going to
> be a security hole.

If you're considered the owner of an object then you have access to drop
it already. You have to be a member of the role to which you're
changing the ownership. That role not having permission to create the
object in place is an interesting question. That's an issue for SET
ROLE too, to some extent I think, do you still have your role's
permissions after you've SET ROLE to another role? If not then you'd
have to grant CREATE on the schema to the role in order to create
objects owned by that role, and I don't think that's necessairly
something you'd want to do.

Thanks,

Stephen

Stephen Frost wrote:
> I can perhaps see a special case for SECURITY DEFINER functions but if
> we're going to special case them I'd think we'd need to make them only
> be creatable/modifiable at all by superusers or add another flag to the
> role to allow that.

I agree that owner changes of SECURITY DEFINER functions seem dangerous. I
would follow Stephen's idea that SECURITY DEFINER functions should only be
creatable/modifiable by superusers.

This would be similar to unix, where setting the suid/sgid bits is usually
only allowed to root.

Best Regards,
Michael Paesold

Stephen Frost wrote:
> If you're considered the owner of an object then you have access to drop
> it already. You have to be a member of the role to which you're
> changing the ownership. That role not having permission to create the
> object in place is an interesting question. That's an issue for SET
> ROLE too, to some extent I think, do you still have your role's
> permissions after you've SET ROLE to another role?

For me this would be the "natural" way how SET ROLE would behave. This is
unix'ism again, but using setuid to become another user, you loose the
privileges of the old user context.
Therefore SET ROLE should not inherit privileges from the other role. This
seems to be the safes approach.

Nevertheless, what does the standard say?

> If not then you'd
> have to grant CREATE on the schema to the role in order to create
> objects owned by that role, and I don't think that's necessairly
> something you'd want to do.

Right, that's an issue. But since the new role will be the *owner* of the
object, it *should* really have create-privileges in that schema. So the
above way seems to be correct anyway.

Best Regards,
Michael Paesold

* Bruno Wolff III (bruno(at)wolff(dot)to) wrote:
> Thinking about it some more, drops wouldn't be an issue since the owner
> can always drop objects.

Right.

> Creating objects in particular schemas or databases is not something that
> all roles may be able to do.

Yeah, I'm not entirely sure what I think about this issue. If you're
not allowed to change ownership of objects and SET ROLE drops your
regular ROLE's privileges then the role which owns the object originally
(and which you're required to be in) must have had create access to that
schema at some point.

I can see requiring the role that's changing the ownership to have
create access to the schema in which the object that's being changed is
in.

Thanks,

Stephen

* Michael Paesold (mpaesold(at)gmx(dot)at) wrote:
> Stephen Frost wrote:
> >If you're considered the owner of an object then you have access to drop
> >it already. You have to be a member of the role to which you're
> >changing the ownership. That role not having permission to create the
> >object in place is an interesting question. That's an issue for SET
> >ROLE too, to some extent I think, do you still have your role's
> >permissions after you've SET ROLE to another role?
>
> For me this would be the "natural" way how SET ROLE would behave. This is
> unix'ism again, but using setuid to become another user, you loose the
> privileges of the old user context.
> Therefore SET ROLE should not inherit privileges from the other role. This
> seems to be the safes approach.
>
> Nevertheless, what does the standard say?

Hmm, it says there's a stack and that the thing on top is what's
currently used, so it sounds like it would drop the privs too, but imv
it's not entirely clear.

> >If not then you'd
> >have to grant CREATE on the schema to the role in order to create
> >objects owned by that role, and I don't think that's necessairly
> >something you'd want to do.
>
> Right, that's an issue. But since the new role will be the *owner* of the
> object, it *should* really have create-privileges in that schema. So the
> above way seems to be correct anyway.

I'm not entirely sure that you'd necessairly want the role to have
create privileges on the schema even when it owns things in the schema
but the more I think about it that doesn't seem all that unreasonable
either. I don't think it'd be very difficult to add such a check to the
ALTER OWNER code too though.

In general, and perhaps as a unix'ism to some extent, I don't
particularly like having to su to people. To get all the other
permissions which the role has you don't have to 'su' currently, and
personally I like that and think that's correct for a role-based
environment (unlike unix where you have users and groups).

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> * Bruno Wolff III (bruno(at)wolff(dot)to) wrote:
>> Creating objects in particular schemas or databases is not something that
>> all roles may be able to do.

> Yeah, I'm not entirely sure what I think about this issue.

We have a precedent, which is that RENAME checks for create rights.
If you want to lean on the argument that this is just a shortcut for
dropping the object and then recreating it somewhere else, then you
need (a) the right to drop the object --- which is inherent in being
the old owner, and (b) the right to create the new object, which means
that (b1) you can become the role you wish to have owning the object,
and (b2) *as that role* you would have the rights needed to create the
object.

Stephen's original analysis covers (a) and (b1) but not (b2). With (b2)
I'd agree that it's just a useful shortcut.

I don't see a need to treat SECURITY DEFINER functions as
superuser-only. We've had that facility since 7.3 or so and no one
has complained that it's too dangerous.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > * Bruno Wolff III (bruno(at)wolff(dot)to) wrote:
> >> Creating objects in particular schemas or databases is not something that
> >> all roles may be able to do.
>
> > Yeah, I'm not entirely sure what I think about this issue.
>
> We have a precedent, which is that RENAME checks for create rights.

Ah, ok. Precedent is good.

> If you want to lean on the argument that this is just a shortcut for
> dropping the object and then recreating it somewhere else, then you
> need (a) the right to drop the object --- which is inherent in being
> the old owner, and (b) the right to create the new object, which means
> that (b1) you can become the role you wish to have owning the object,
> and (b2) *as that role* you would have the rights needed to create the
> object.
>
> Stephen's original analysis covers (a) and (b1) but not (b2). With (b2)
> I'd agree that it's just a useful shortcut.

Right. Ok, I'll develop a patch which covers (a), (b1) and (b2). I'll
also go through all of the superuser() calls in src/backend/commands/
and check for other places we may need *_ownercheck calls.

I expect to have the patch done either tonight or tommorow.

Thanks,

Stephen

Greetings,

Attached please find a patch to change how the permissions checking
for alter-owner is done. With roles there can be more than one
'owner' of an object and therefore it becomes sensible to allow
specific cases of ownership change for non-superusers.

The permission checks for change-owner follow the alter-rename
precedent that the new owner must have permission to create the object
in the schema.

The roles patch previously applied did not require the role for
which a database is being created to have createdb privileges, or for
the role for which a schema is being created to have create
privileges on the database (the role doing the creation did have to
have those privileges though, of course).

For 'container' type objects this seems reasonable. 'container' type
objects are unlike others in a few ways, but one of the more notable
differences for this case is that an owner may be specified as part of
the create command.

To support cleaning up the various checks, I also went ahead and
modified is_member_of_role() to always return true when asked if a
superuser is in a given role. This seems reasonable, won't affect
what's actually seen in the various tables, and allows us to eliminate
explicit superuser() checks in a number of places.

I have also reviewed the other superuser() calls in
src/backend/commands/ and feel pretty comfortable that they're all
necessary, reasonable, and don't need to be replaced with
*_ownercheck or other calls.

The specific changes which have been changed, by file:
aggregatecmds.c, alter-owner:
alter-owner checks:
User is owner of the to-be-changed object
User is a member of the new owner's role
New owner is permitted to create objects in the schema
Superuser() requirement removed

conversioncmds.c, rename:
rename-checks:
Changed from superuser() or same-roleId to pg_conversion_ownercheck
alter-owner checks:
User is owner of the to-be-changed object
User is a member of the new owner's role
New owner is permitted to create objects in the schema
Superuser() requirement removed

dbcommands.c:
Moved superuser() check to have_createdb_privilege
Cleaned up permissions checking in createdb and rename
alter-owner checks:
User is owner of the database
User is a member of the new owner's role
User has createdb privilege

functioncmds.c:
alter-owner checks:
User is owner of the function
User is a member of the new owner's role
New owner is permitted to create objects in the schema

opclasscmds.c:
alter-owner checks:
User is owner of the object
User is a member of the new owner's role
New owner has permission to create objects in the schema

operatorcmds.c:
alter-owner checks:
User is owner of the object
User is a member of the new owner's role
New owner has permission to create objects in the schema

schemacmds.c:
Cleaned up create schema identify changing/setting/checking
(This code was quite different from all the other create functions,
these changes make it much more closely match createdb)
alter-owner checks:
User is owner of the schema
User is a member of the new owner's role
User has create privilege on database

tablecmds.c:
alter-owner checks:
User is owner of the object
User is a member of the new owner's role
New owner has permission to create objects in the schema

tablespace.c:
alter-owner checks:
User is owner of the tablespace
User is a member of the new owner's role
(No create-tablespace permission to check, tablespaces must be
created by superusers and so alter-owner here really only matters
if the superuser changed the tablespace owner to a non-superuser
and then that non-superuser wants to change the ownership to yet
another user, the other option would be to continue to force
superuser-only for tablespace owner changes but I'm not sure I
see the point if the superuser trusts the non-superuser enough to
give them a tablespace...)

typecmds.c:
alter-owner checks:
User is owner of the object
User is a member of the new owner's role
New owner has permission to create objects in the schema

Many thanks. As always, comments, questions, concerns, please let me
know.

Thanks again,

Stephen

I notice that AddRoleMems/DelRoleMems assume that ADMIN OPTION is not
inherited indirectly; that is it must be granted directly to you.
This seems wrong; SQL99 has under <privileges>

19) B has the WITH ADMIN OPTION on a role if a role authorization
descriptor identifies the role as granted to B WITH ADMIN OPTION
or a role authorization descriptor identifies it as granted WITH
ADMIN OPTION to another applicable role for B.

and in the Access Rules for <grant role statement>

1) Every role identified by <role granted> shall be contained
in the applicable roles for A and the corresponding role
authorization descriptors shall specify WITH ADMIN OPTION.

I can't see any support in the spec for the idea that WITH ADMIN OPTION
doesn't flow through role memberships in the same way as ordinary
membership; can you quote someplace that implies this?

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> I notice that AddRoleMems/DelRoleMems assume that ADMIN OPTION is not
> inherited indirectly; that is it must be granted directly to you.
> This seems wrong; SQL99 has under <privileges>
>
> 19) B has the WITH ADMIN OPTION on a role if a role authorization
> descriptor identifies the role as granted to B WITH ADMIN OPTION
> or a role authorization descriptor identifies it as granted WITH
> ADMIN OPTION to another applicable role for B.
>
> and in the Access Rules for <grant role statement>
>
> 1) Every role identified by <role granted> shall be contained
> in the applicable roles for A and the corresponding role
> authorization descriptors shall specify WITH ADMIN OPTION.
>
> I can't see any support in the spec for the idea that WITH ADMIN OPTION
> doesn't flow through role memberships in the same way as ordinary
> membership; can you quote someplace that implies this?

Hrm, no, sorry, I just interpreted the 'Access Rules' line for <grant
role statement> differently. That is to say:

1) Every role identified by <role granted> shall be contained
(Alright, all the roles which you're granting, right)

in the applicable roles for A and the corresponding role
(A must be in all the roles which are being granted)

authorization descriptors shall specify WITH ADMIN OPTION.
(the grants to A for those rules specify ADMIN OPTION)

This came across to me as meaning "there must exist an authorization
descriptor such that the granted-role equals <role granted>, the grantee
is A and WITH ADMIN OPTION is set". That could only be true if the
grant was done explicitly. Reading from 19 above (which I don't recall
seeing before, or at least not reading very carefully) I think you're
right. Either way is fine with me.

Thanks,

Stephen

Dear Stephen,

> Right, it's a beginning to proper 'Role' support as defined by the SQL
> specification.

Ok. AFAIC remember, the specification is pretty subtle and fuzzy enough so
that there is room for little design options.

> I understand your concerns here and while I agree with the basic idea
> that per-catalog role sets would be nice it wasn't what I had set out to
> do with this patch. Perhaps what you're asking for will be added later
> on.
>
> Some things this patch does do though are:
> [...]

Ok. I think I understand but I'm not sure it is done the right way.
Let me explain my (possibly wrong) point of view:

The standard talks about 2 distinct concepts: USER and ROLE (4.34). I'm
not sure it is a good idea to drop the user concept to replace it by role.
If you do so, you may miss something about what roles are about.

The SESSION_USER/CURRENT_USER has a CURRENT_ROLE which defines the rights
at a given time. This role can be changed by the user, based on user/role
membership, so the user can change its 'effective' rights. Roles are
potential privileges that a user can set himself in if he/she desires so.
example:

sh> psql -u calvin mydb
calvin(at)mydb>
-- I'm user calvin with no role or a default role on mydb.
-- I can do all which is allowed to 'calvin' as a user.

calvin(at)mydb> SET ROLE admin;
calvin/admin(at)mydb> ...
-- I'm allowed to do that if the role 'admin' is granted to 'calvin'
-- now I can do whatever is allowed to role 'admin'.

calvin/admin(at)mydb> SET ROLE basic;
calvin/basic(at)mydb> ...
-- now I can do what is allowed to role 'basic' and the roles 'basic' are in.
-- things that where allowed to admin may *not* be accessible now.

This is a very useful feature, and a key idea of the specs IMVVHO. ISTM
that the way "fuse" user and role misses that important point, as I have
not seen a "set role" in the grammar file.

Although in the spec role rights are transitive in the role realm, it
should *stop* at the user. If you drop the user concept, you just have a
group with automatically provided rights.

The fact that the spec does not specify the USER stuff and specifies the
ROLE stuff does not mean that having only roles is the good way to go.

So for me we should have per-cluser users as they where up to now,
per-catalog roles with the properties I described, and possibly
per-cluster group just for the sake of compatibility/simplicity of the
access control and managing group of users as a whole. ROLE should not
replace USER/GROUP. It should be added next to it.

Maybe I'm wrong at my reading of the spec and its intent, and at my quick
check through the status of the cvs head, but that's my current
understanding, and I think it should be checked seriously.

Have a nice day,

--
Fabien

Fabien,

* Fabien COELHO (fabien(at)coelho(dot)net) wrote:
> This is a very useful feature, and a key idea of the specs IMVVHO. ISTM
> that the way "fuse" user and role misses that important point, as I have
> not seen a "set role" in the grammar file.

'set role' is coming, sorry it wasn't in my initial patch. We are
looking to support pretty much all of the SQL features 'Basic roles' and
'Extended roles'. I think 'set role' and maybe some cleanup of
information_schema is all we need to claim 'Basic roles' support. For
'Extended roles' I think we need revoke role cascade support.
Interestingly, the SQL2003 draft I'm looking at doesn't list 'drop role
<role name> cascade' as being valid, yet other places in the spec
specify 'drop role <role name> cascade' usage, so I think we should
support that.

> Although in the spec role rights are transitive in the role realm, it
> should *stop* at the user. If you drop the user concept, you just have a
> group with automatically provided rights.

I'm not quite sure what you mean here. Role right resolution starts
from the user and then works backwards up the tree, with multi-level
resolution. It wouldn't go past the logged in user since that's really
where it starts.

> The fact that the spec does not specify the USER stuff and specifies the
> ROLE stuff does not mean that having only roles is the good way to go.

I'm pretty sure we'll be able to match the SQL spec and support at least
everything we did before with users/groups...

> So for me we should have per-cluser users as they where up to now,
> per-catalog roles with the properties I described, and possibly
> per-cluster group just for the sake of compatibility/simplicity of the
> access control and managing group of users as a whole. ROLE should not
> replace USER/GROUP. It should be added next to it.

I don't see much point in having USER or GROUP when we have roles. Is
there something specific that you feel can't be done with roles that
could be done w/ USER/GROUP? Per-catalog roles is an interesting idea,
but I'd tend to think that if you want per-catalog roles, you'd want
per-catalog users too. I don't have any problem with that, but I don't
see how not being per-catalog indicates we should have USER/GROUP
instead of roles.

> Maybe I'm wrong at my reading of the spec and its intent, and at my quick
> check through the status of the cvs head, but that's my current
> understanding, and I think it should be checked seriously.

I just went through the spec yesterday, check -hackers for my email
about what CVS head supports vs. what's in the SQL spec. I don't see
any particular reason why we wouldn't be able to fully support 'Basic
roles' and 'Extended roles' in 8.1, I think we're quite close now...

Thanks,

Stephen

Dear Stephen,

Thanks again on working on this feature.

> Role right resolution starts from the user and then works backwards up
> the tree, with multi-level resolution. It wouldn't go past the logged
> in user since that's really where it starts.

ISTM that the starting point should *not* be the user, but the
CURRENT_ROLE, which must be something distinct: Even if I'm root, if a
'SET ROLE very_limited_privileges' is performed, then the privileges in
effect are those of the chosen role. That is what is told by section
4.34.1.1 "SQL-session authorization identifiers" of the SQL 2003 specs as
I understand it.

If the user is kind of a role, then I'm afraid the whole point may be
missed. But maybe not, it would depend on the implementation details.

>> So for me we should have per-cluser users as they where up to now,
>> per-catalog roles with the properties I described, and possibly
>> per-cluster group just for the sake of compatibility/simplicity of the
>> access control and managing group of users as a whole. ROLE should not
>> replace USER/GROUP. It should be added next to it.
>
> I don't see much point in having USER or GROUP when we have roles.

Indeed, if you have per-cluster ROLE, you don't need GROUP anymore.

If USER is per-cluster for connection management and ROLE per-catalog for
database access management, then you will need a per-cluster grouping
(say for pg_hba.conf...) which is just the current GROUP.

> Is there something specific that you feel can't be done with roles that
> could be done w/ USER/GROUP?

No, it is the reverse: I'm afraid that the way it seems to be heading, no
more will be done with role than with group before.

> Per-catalog roles is an interesting idea, but I'd tend to think that if
> you want per-catalog roles, you'd want per-catalog users too.

I'm fine with per-cluster users.

> I just went through the spec yesterday, check -hackers for my email

Ok, I'm going to look into that.

> about what CVS head supports vs. what's in the SQL spec. I don't see
> any particular reason why we wouldn't be able to fully support 'Basic
> roles' and 'Extended roles' in 8.1, I think we're quite close now...

I'm looking forward to the 'SET ROLE' implementation. If the
interpretation of the privileges is restricted to the current role, then
I'll be happy.

I still think that removing groups and having per-cluster roles is not a
good idea. The better way would be to keep user/group and add per-catalog
roles. There is an opportunity which is being missed, and that won't show
up later. Well, I can see that I'm pretty alone to think that;-)

Thanks for your answer, have a nice day,

--
Fabien.

Fabien & Tom (if you're watching),

* Fabien COELHO (coelho(at)cri(dot)ensmp(dot)fr) wrote:
> >Role right resolution starts from the user and then works backwards up
> >the tree, with multi-level resolution. It wouldn't go past the logged
> >in user since that's really where it starts.
>
> ISTM that the starting point should *not* be the user, but the
> CURRENT_ROLE, which must be something distinct: Even if I'm root, if a

Err, yes, from the CURRENT_ROLE, which is what we get back from
'GetUserId()', technically I think. I'm not sure the stack has really
been implemented yet (I'd expect it to be done correctly w/ the SET ROLE
things).

> 'SET ROLE very_limited_privileges' is performed, then the privileges in
> effect are those of the chosen role. That is what is told by section
> 4.34.1.1 "SQL-session authorization identifiers" of the SQL 2003 specs as
> I understand it.

Right, that's what the behavior should be.

> If the user is kind of a role, then I'm afraid the whole point may be
> missed. But maybe not, it would depend on the implementation details.

No, I don't think the point will be missed at all. I certainly
understand that privileges are dropped when doing SET ROLE. Really, I
think SET ROLE and the associated SESSION_USER/CURRENT_USER/CURRENT_ROLE
need to be implemented/looked at carefully to make sure the right things
happen.

Tom, if you're watching, are you working on this? I can probably spend
some time today on it, if that'd be helpful.

> >>So for me we should have per-cluser users as they where up to now,
> >>per-catalog roles with the properties I described, and possibly
> >>per-cluster group just for the sake of compatibility/simplicity of the
> >>access control and managing group of users as a whole. ROLE should not
> >>replace USER/GROUP. It should be added next to it.
> >
> >I don't see much point in having USER or GROUP when we have roles.
>
> Indeed, if you have per-cluster ROLE, you don't need GROUP anymore.
>
> If USER is per-cluster for connection management and ROLE per-catalog for
> database access management, then you will need a per-cluster grouping
> (say for pg_hba.conf...) which is just the current GROUP.
>
> >Is there something specific that you feel can't be done with roles that
> >could be done w/ USER/GROUP?
>
> No, it is the reverse: I'm afraid that the way it seems to be heading, no
> more will be done with role than with group before.

Already at least some of the things I was looking for with Roles can be
done, such as a role with members having ownership of an object- this
allows me to create 'admin' roles for a given area without having to
give them superuser(). It's not perfect yet, but it's getting closer.

> >Per-catalog roles is an interesting idea, but I'd tend to think that if
> >you want per-catalog roles, you'd want per-catalog users too.
>
> I'm fine with per-cluster users.

I'm pretty sure others have been asking about per-catalog users and if
we're going to accept that per-catalog roles makes sense I'd really
think per-catalog users would too.

> >about what CVS head supports vs. what's in the SQL spec. I don't see
> >any particular reason why we wouldn't be able to fully support 'Basic
> >roles' and 'Extended roles' in 8.1, I think we're quite close now...
>
> I'm looking forward to the 'SET ROLE' implementation. If the
> interpretation of the privileges is restricted to the current role, then
> I'll be happy.

Right, according to the SQL spec it's a 'stack', where generally the
only thing visible, and what's used for permissions checking, etc, is
whatever is on the top of the stack, so after a 'SET ROLE' you only have
the permissions of that 'SET ROLE'. The only concern I can see here is
that I'm pretty sure the SQL spec allows you to go back (Using 'SET
ROLE' with no argument, or maybe 'SET ROLE NONE', I'd have to
double-check). That makes sense in some instances, but not in others.
There might be room to consider something like 'SET ROLE <role> FINAL'
or some such which disallows going back, though that'd be a PG extension
beyond the SQL spec I'm pretty sure.

> I still think that removing groups and having per-cluster roles is not a
> good idea. The better way would be to keep user/group and add per-catalog
> roles. There is an opportunity which is being missed, and that won't show
> up later. Well, I can see that I'm pretty alone to think that;-)

I really disagree with you here. I feel it makes much more sense to do
this in stages, first user/group -> roles, then roles-per-catalog, which
means you can then have both per-catalog 'users' and per-catalog
'groups', if you want to limit your view to that.

> Thanks for your answer, have a nice day,

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> Tom, if you're watching, are you working on this? I can probably spend
> some time today on it, if that'd be helpful.

I am not; I was hoping you'd deal with SET ROLE. Is it really much
different from SET SESSION AUTHORIZATION?

> I'm pretty sure others have been asking about per-catalog users and if
> we're going to accept that per-catalog roles makes sense I'd really
> think per-catalog users would too.

We really can't do this. Especially not 3 days before feature freeze.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > Tom, if you're watching, are you working on this? I can probably spend
> > some time today on it, if that'd be helpful.
>
> I am not; I was hoping you'd deal with SET ROLE. Is it really much
> different from SET SESSION AUTHORIZATION?

No, it's not, but it's going to need to be done carefully to make sure
GetUserId() returns the correct thing at the correct time and that the
other GetSessionUserId() calls are only used where they should be and
that they return the correct information too.

I'll work on SET ROLE and the associated CURRENT_* functions and
information_schema today and tommorow.

> > I'm pretty sure others have been asking about per-catalog users and if
> > we're going to accept that per-catalog roles makes sense I'd really
> > think per-catalog users would too.
>
> We really can't do this. Especially not 3 days before feature freeze.

Right, I wasn't expecting that to be done in this round. It's something
people have asked for though and so might be something to consider for
8.2. I'm hoping your work on CREATEROLE will stem some of that demand
for per-catalog users/roles actually. I've been trying to think what
else per-catalog users/roles would get us besides a segmented namespace.
I think one big issue is that we don't have a 'usage' database check
beyond pg_hba and so any user could get the schema definitions for any
database, which kind of sucks. Is that maybe something we could try to
address for 8.1?

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> I think one big issue is that we don't have a 'usage' database check
> beyond pg_hba and so any user could get the schema definitions for any
> database, which kind of sucks.

Not unless he can connect to it.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > I think one big issue is that we don't have a 'usage' database check
> > beyond pg_hba and so any user could get the schema definitions for any
> > database, which kind of sucks.
>
> Not unless he can connect to it.

That's controlled by pg_hba.conf though, isn't it? The idea being that
you'd like to give some people the ability to create users/roles, but to
limit the databases those created users/roles could connect to by, say,
requiring they have 'usage' or 'connect' permissions to that database,
which could be set by the database owner; without the database owner
having write permissions to the pg_hba.conf.

The scenario is one of an ISP who wants to give people Postgres access
but doesn't want to have to manage all the users. So, the ISP creates
a database, an 'admin' role for a given customer and gives 'createrole'
permissions to that admin role. The admin role can then create new
roles but can only give them access to connect to their database (since
that's the only one the admin either 'owns' or has 'create', etc,
privileges on). I *think* (perhaps I'm wrong..) that the only thing we
lack to make this work is a permissions check on the connect to a given
database which can be managed by a user of the database (ie: not
pg_hba.conf).

Thinking about this a bit more I guess this would probably involve
basically moving pg_hba.conf into the database catalogs and then having
pg_hba.conf generated similar to how pg_authid is generated. That's
probably too much to do for 8.1 then, I had been hoping there was a way
to do it which would be a smaller change than that.

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> That's controlled by pg_hba.conf though, isn't it? The idea being that
> you'd like to give some people the ability to create users/roles, but to
> limit the databases those created users/roles could connect to by, say,
> requiring they have 'usage' or 'connect' permissions to that database,
> which could be set by the database owner; without the database owner
> having write permissions to the pg_hba.conf.

You can do that today by putting a group name in pg_hba.conf. Roles
will make it more flexible; I don't see that we need anything more.

For instance, if pg_hba.conf says "samegroup" then you could manage
everything by associating a group with each database.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > That's controlled by pg_hba.conf though, isn't it? The idea being that
> > you'd like to give some people the ability to create users/roles, but to
> > limit the databases those created users/roles could connect to by, say,
> > requiring they have 'usage' or 'connect' permissions to that database,
> > which could be set by the database owner; without the database owner
> > having write permissions to the pg_hba.conf.
>
> You can do that today by putting a group name in pg_hba.conf. Roles
> will make it more flexible; I don't see that we need anything more.
>
> For instance, if pg_hba.conf says "samegroup" then you could manage
> everything by associating a group with each database.

Ahh, ok, good point. Sorry, I'd forgotten about that flexibility of
pg_hba.conf. Well, hopefully this will make some ISPs happy then. :)

Thanks,

Stephen

Dear Stephen,

>> I still think that removing groups and having per-cluster roles is not a
>> good idea. The better way would be to keep user/group and add per-catalog
>> roles. There is an opportunity which is being missed, and that won't show
>> up later.
>
> I really disagree with you here. I feel it makes much more sense to do
> this in stages, first user/group -> roles, then roles-per-catalog, which
> means you can then have both per-catalog 'users' and per-catalog
> 'groups', if you want to limit your view to that.

I don't think that having two kinds of roles (per-cluster and per-catalog)
would be a practical thing from the user perspective. From the
implementation point of view, two tables will be needed. If you don't
create roles directly in the right scope, it will create confusion later.

The two concept need to have two different names so that they can be
understood. Moreover, a working per-cluster grouping was already
available. Changing the role scope will be much harder than creating a
role directly in the good scope.

From the implementation perspective, there is more work at adding
per-cluster roles and removing per-cluster group and then later try to add
per-catalog roles than adding per-catalog roles directly without touching
the existing group stuff.

So I'm afraid that the opportunity is missed and that per-catalog role
will never get in. Well, at least you look more optimistic than me;-)

--
Fabien.

Fabien,

* Fabien COELHO (coelho(at)cri(dot)ensmp(dot)fr) wrote:
> >I really disagree with you here. I feel it makes much more sense to do
> >this in stages, first user/group -> roles, then roles-per-catalog, which
> >means you can then have both per-catalog 'users' and per-catalog
> >'groups', if you want to limit your view to that.
>
> I don't think that having two kinds of roles (per-cluster and per-catalog)
> would be a practical thing from the user perspective. From the
> implementation point of view, two tables will be needed. If you don't
> create roles directly in the right scope, it will create confusion later.
>
> The two concept need to have two different names so that they can be
> understood. Moreover, a working per-cluster grouping was already
> available. Changing the role scope will be much harder than creating a
> role directly in the good scope.

The two concepts certainly don't need different names to distinguish
them. A simple distinction such as superusers are per-cluster and all
other roles are not would be sufficient. I expect that's the kind of
thing people would be looking for anyway.

> >From the implementation perspective, there is more work at adding
> per-cluster roles and removing per-cluster group and then later try to add
> per-catalog roles than adding per-catalog roles directly without touching
> the existing group stuff.

Having just spent a fair bit of time on the implementation, I have to
disagree with you here.

> So I'm afraid that the opportunity is missed and that per-catalog role
> will never get in. Well, at least you look more optimistic than me;-)

Honestly, this comes across to me the same as saying that because we
have databases we'd never have schemas.

Please outline exactly what you're really looking for. Let's drop the
idea of per-cluster users/groups/roles/whatever and instead consider
what specific capabilities you're looking for. We can then decide if
those capabilities are best provided through per-catalog roles, if
they're already covered with the existing framework, or if there's some
other, better solution.

Thanks,

Stephen

Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr> writes:
> Moreover, a working per-cluster grouping was already
> available.

Only for sufficiently small values of "working". The lack of ability
for groups to contain other groups and for groups to be the direct
owners of objects were both pretty serious restrictions, which are now
fixed.

regards, tom lane

Dear Stephen,

> Please outline exactly what you're really looking for. Let's drop the
> idea of per-cluster users/groups/roles/whatever and instead consider
> what specific capabilities you're looking for.

I think from a conceptual point of view that the ability to manage
permissions at the database level (per-catalog role) is a good thing (tm)
for everybody. The privilege management is about a catalog, so it better
to have it in the catalog.

My personnal uses are two fold :

- for teaching purposes, I can give every student his/her database
and have her/him practice db privileges independently. They
can create their own roles and do whatever with them...

- for administration purposes, different databases have different
requirements, and maybe different kind of role "readonly",
"modifiable", "fulladmin" which could be attributed independently in
each database.

Basically, I want to be able to delegate to someone the right management
for one database, including the creation of roles and so on, without
interference from one database to another.

So as to illustrate what I call an interference: say you have 2 databases
which where on 2 clusters and you want to transfert them into the same
cluster. If the same role name was used in both database, you may have
interferences, people being given rights on one database and applying them
to the other if they can connect to it.

> We can then decide if those capabilities are best provided through
> per-catalog roles, if they're already covered with the existing
> framework, or if there's some other, better solution.

One inelegant solution is to prefix the role names with the database name,
but that is just a discipline and cannot be inforced. I think we can do better.

If you're right that having both "per-catalog" and "per-cluster" role with
some flag would be accepted into postgresql, then that will be fine with
me. I'll just argue for the per-catalog roles to be the default.

Thanks for all your answers, have a nice day,

--
Fabien.

Fabien,

* Fabien COELHO (coelho(at)cri(dot)ensmp(dot)fr) wrote:
> >Please outline exactly what you're really looking for. Let's drop the
> >idea of per-cluster users/groups/roles/whatever and instead consider
> >what specific capabilities you're looking for.
>
> I think from a conceptual point of view that the ability to manage
> permissions at the database level (per-catalog role) is a good thing (tm)
> for everybody. The privilege management is about a catalog, so it better
> to have it in the catalog.

Permissions are at a number of levels already: cluster, database,
schema, table. Permissions at different levels hasn't got anything to
do w/ per-catalog roles.

> My personnal uses are two fold :
>
> - for teaching purposes, I can give every student his/her database
> and have her/him practice db privileges independently. They
> can create their own roles and do whatever with them...

Right, this can be done now. When you set up each student with his/her
database create two roles:
Role 1: Has createrole permissions and admin permissions on role 2.
Role 2: Add into pg_hba.conf as a role with permission to access the db.

When a student wants to create another role with access to the db they
just have to log in as Role 1 and create the role and add that role to
Role 2. That role can then log in to their database.

> - for administration purposes, different databases have different
> requirements, and maybe different kind of role "readonly",
> "modifiable", "fulladmin" which could be attributed independently in
> each database.

I don't see how this has got anything to do w/ per-catalog roles
either...

> Basically, I want to be able to delegate to someone the right management
> for one database, including the creation of roles and so on, without
> interference from one database to another.

That's what createrole should let you do w/ current CVS HEAD. Don't
thank me though, Tom did the heavy lifting wrt that.

> So as to illustrate what I call an interference: say you have 2 databases
> which where on 2 clusters and you want to transfert them into the same
> cluster. If the same role name was used in both database, you may have
> interferences, people being given rights on one database and applying them
> to the other if they can connect to it.

Ah-hah, now here's something we can talk about: namespace collision.
That's an issue which per-catalog roles would help with. I'm not so
sure that'd make administration *easier* though, I'd think it'd make it
harder, honestly, at least in the case where you've got multiple
databases that you want to give a certain person access to.

> >We can then decide if those capabilities are best provided through
> >per-catalog roles, if they're already covered with the existing
> >framework, or if there's some other, better solution.
>
> One inelegant solution is to prefix the role names with the database name,
> but that is just a discipline and cannot be inforced. I think we can do
> better.

That's essentially all you're really asking for though, and is something
which could be done in the current framework. It'd probably be more
elegant to have a per-catalog pg_authid though. As long as we can
identify the database trying to be connected to at the same time or
before we get the username I don't think this would be too hard to
support. Perhaps for 8.2 this could be done, though I'm still not
convinced it's a definite win.

> If you're right that having both "per-catalog" and "per-cluster" role with
> some flag would be accepted into postgresql, then that will be fine with
> me. I'll just argue for the per-catalog roles to be the default.

It'd really be a create-role option, 'create role abc GLOBAL' or some
such. The resolution would then be check the per-catalog pg_authid
first and if nothing is found there go to the system-wide pg_authid.
That's kind of ugly, of course, and could slow things down for people
who prefer to have global roles instead of per-catalog roles.

Thanks,

Stephen

>> The privilege management is about a catalog, so it better to have it in
>> the catalog.
>
> Permissions are at a number of levels already: cluster, database,
> schema, table. Permissions at different levels hasn't got anything to
> do w/ per-catalog roles.

Sorry for not being very clear. I see two "main" levels:

(1) the connection which is managed in pg_hba.conf. It is a sysadmin
concern, where you decide who will be able to get into your system. This
issue is *not* addressed by the SQL specs.

(2) once you're connected to a catalog, the ability to adjust/organize
privileges for accessing data within this catalog. This is a dbadmin
concern, where you decide for a given user which can access the database,
what data should be readable. This issue is addressed by the SQL specs.

If you think unix, root decides the first, and each user decides the
second for the files it owns.

>> - for teaching purposes [...]
>
> Right, this can be done now.

There is the namespace collision issue, and although I might grant a
student the privilege to create simple roles, I would not allow them to
create new users for a basic practice;-)

>> - for administration purposes, different databases have different
>> requirements, and maybe different kind of role "readonly",
>> "modifiable", "fulladmin" which could be attributed independently in
>> each database.
>
> I don't see how this has got anything to do w/ per-catalog roles either...

Because of namespace collision. That what I mean by "independently" above.

>> So as to illustrate what I call an interference [...]
>
> Ah-hah, now here's something we can talk about: namespace collision.

That is just what I meant in the last 10 mails when I mention per-catalog
roles as better than per-cluster roles. I just did not put in the right
keywords to make myself clear:-( Sigh.

> That's an issue which per-catalog roles would help with.

Indeed.

> I'm not so sure that'd make administration *easier* though, I'd think
> it'd make it harder, honestly, at least in the case where you've got
> multiple databases that you want to give a certain person access to.

Sure, "grouping" at the cluster level is also useful.

> [...]
> As long as we can identify the database trying to be connected to at the
> same time or before we get the username I don't think this would be too
> hard to support.

Yet, but this is not what I'm looking for. I want the grouping
per-catalog, but the user (or connectable-role now) is fine enough for me
at the cluster level.

> Perhaps for 8.2 this could be done, though I'm still not convinced it's
> a definite win.

For the "user per-catalog" part, I agree with you.

>> If you're right that having both "per-catalog" and "per-cluster" role with
>> some flag would be accepted into postgresql, then that will be fine with
>> me. I'll just argue for the per-catalog roles to be the default.
>
> It'd really be a create-role option, 'create role abc GLOBAL'

There is also a problem of namespace collision if you have both global and
local roles. When I create a global role, pg cannot check that this name
is not used in ANY of the databases. If you can have two "admin" roles,
one global and one local... I doubt Tom will let pass such an improvement.
Also, I don't feel the upward compatibility constraint well with
per-catalog roles once per-cluster roles are in place.

> or some such. The resolution would then be check the per-catalog
> pg_authid first and if nothing is found there go to the system-wide
> pg_authid. That's kind of ugly, of course, and could slow things down
> for people who prefer to have global roles instead of per-catalog roles.

If the per-catalog role is empty, I guess an obvious optimisation could be
implemented;-)

Good night,

--
Fabien.

Stupid question, but how do roles relate to our existing "groups"?

---------------------------------------------------------------------------

Fabien COELHO wrote:
>
> >> The privilege management is about a catalog, so it better to have it in
> >> the catalog.
> >
> > Permissions are at a number of levels already: cluster, database,
> > schema, table. Permissions at different levels hasn't got anything to
> > do w/ per-catalog roles.
>
> Sorry for not being very clear. I see two "main" levels:
>
> (1) the connection which is managed in pg_hba.conf. It is a sysadmin
> concern, where you decide who will be able to get into your system. This
> issue is *not* addressed by the SQL specs.
>
> (2) once you're connected to a catalog, the ability to adjust/organize
> privileges for accessing data within this catalog. This is a dbadmin
> concern, where you decide for a given user which can access the database,
> what data should be readable. This issue is addressed by the SQL specs.
>
> If you think unix, root decides the first, and each user decides the
> second for the files it owns.
>
> >> - for teaching purposes [...]
> >
> > Right, this can be done now.
>
> There is the namespace collision issue, and although I might grant a
> student the privilege to create simple roles, I would not allow them to
> create new users for a basic practice;-)
>
> >> - for administration purposes, different databases have different
> >> requirements, and maybe different kind of role "readonly",
> >> "modifiable", "fulladmin" which could be attributed independently in
> >> each database.
> >
> > I don't see how this has got anything to do w/ per-catalog roles either...
>
> Because of namespace collision. That what I mean by "independently" above.
>
> >> So as to illustrate what I call an interference [...]
> >
> > Ah-hah, now here's something we can talk about: namespace collision.
>
> That is just what I meant in the last 10 mails when I mention per-catalog
> roles as better than per-cluster roles. I just did not put in the right
> keywords to make myself clear:-( Sigh.
>
> > That's an issue which per-catalog roles would help with.
>
> Indeed.
>
> > I'm not so sure that'd make administration *easier* though, I'd think
> > it'd make it harder, honestly, at least in the case where you've got
> > multiple databases that you want to give a certain person access to.
>
> Sure, "grouping" at the cluster level is also useful.
>
> > [...]
> > As long as we can identify the database trying to be connected to at the
> > same time or before we get the username I don't think this would be too
> > hard to support.
>
> Yet, but this is not what I'm looking for. I want the grouping
> per-catalog, but the user (or connectable-role now) is fine enough for me
> at the cluster level.
>
> > Perhaps for 8.2 this could be done, though I'm still not convinced it's
> > a definite win.
>
> For the "user per-catalog" part, I agree with you.
>
> >> If you're right that having both "per-catalog" and "per-cluster" role with
> >> some flag would be accepted into postgresql, then that will be fine with
> >> me. I'll just argue for the per-catalog roles to be the default.
> >
> > It'd really be a create-role option, 'create role abc GLOBAL'
>
> There is also a problem of namespace collision if you have both global and
> local roles. When I create a global role, pg cannot check that this name
> is not used in ANY of the databases. If you can have two "admin" roles,
> one global and one local... I doubt Tom will let pass such an improvement.
> Also, I don't feel the upward compatibility constraint well with
> per-catalog roles once per-cluster roles are in place.
>
> > or some such. The resolution would then be check the per-catalog
> > pg_authid first and if nothing is found there go to the system-wide
> > pg_authid. That's kind of ugly, of course, and could slow things down
> > for people who prefer to have global roles instead of per-catalog roles.
>
> If the per-catalog role is empty, I guess an obvious optimisation could be
> implemented;-)
>
> Good night,
>
> --
> Fabien.
>
> ---------------------------(end of broadcast)---------------------------
> TIP 3: if posting/reading through Usenet, please send an appropriate
> subscribe-nomail command to majordomo(at)postgresql(dot)org so that your
> message can get through to the mailing list cleanly
>

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073

* Bruce Momjian (pgman(at)candle(dot)pha(dot)pa(dot)us) wrote:
> Stupid question, but how do roles relate to our existing "groups"?

Uhhh. There are no longer "groups", they've been replaced with roles
(which can have members).

Thanks,

Stephen

Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> Stupid question, but how do roles relate to our existing "groups"?

As committed, roles subsume both users and groups: a role that permits
login (rolcanlogin) acts as a user, and a role that has members is a
group. It is possible for the same role to do both things, though I'm
not sure that it's good security policy to set up a role that way.

The advantage over what we had is exactly that there isn't any
distinction, and thus groups can do everything users can and
vice versa:
* groups can own objects
* groups can contain other groups (we forbid loops though)

Also there is a notion of "admin option" for groups, which is like
"grant option" for privileges: you can designate certain members of
a group as being able to grant ownership in that group to others,
without having to make them superusers.

regards, tom lane

Thanks, TODO updated. We still support CREATE GROUP? It translates to
roles?

---------------------------------------------------------------------------

Tom Lane wrote:
> Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > Stupid question, but how do roles relate to our existing "groups"?
>
> As committed, roles subsume both users and groups: a role that permits
> login (rolcanlogin) acts as a user, and a role that has members is a
> group. It is possible for the same role to do both things, though I'm
> not sure that it's good security policy to set up a role that way.
>
> The advantage over what we had is exactly that there isn't any
> distinction, and thus groups can do everything users can and
> vice versa:
> * groups can own objects
> * groups can contain other groups (we forbid loops though)
>
> Also there is a notion of "admin option" for groups, which is like
> "grant option" for privileges: you can designate certain members of
> a group as being able to grant ownership in that group to others,
> without having to make them superusers.
>
> regards, tom lane
>

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073

* Bruce Momjian (pgman(at)candle(dot)pha(dot)pa(dot)us) wrote:
> Thanks, TODO updated. We still support CREATE GROUP? It translates to
> roles?

Yes, CREATE USER too.

Stephen

> Tom Lane wrote:
> > Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us> writes:
> > > Stupid question, but how do roles relate to our existing "groups"?
> >
> > As committed, roles subsume both users and groups: a role that permits
> > login (rolcanlogin) acts as a user, and a role that has members is a
> > group. It is possible for the same role to do both things, though I'm
> > not sure that it's good security policy to set up a role that way.
> >
> > The advantage over what we had is exactly that there isn't any
> > distinction, and thus groups can do everything users can and
> > vice versa:
> > * groups can own objects
> > * groups can contain other groups (we forbid loops though)
> >
> > Also there is a notion of "admin option" for groups, which is like
> > "grant option" for privileges: you can designate certain members of
> > a group as being able to grant ownership in that group to others,
> > without having to make them superusers.
> >
> > regards, tom lane
> >
>
> --
> Bruce Momjian | http://candle.pha.pa.us
> pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
> + If your life is a hard drive, | 13 Roberts Road
> + Christ can be your backup. | Newtown Square, Pennsylvania 19073

On Friday 01 July 2005 13:07, Stephen Frost wrote:
> * Bruce Momjian (pgman(at)candle(dot)pha(dot)pa(dot)us) wrote:
> > Thanks, TODO updated. We still support CREATE GROUP? It translates to
> > roles?
>
> Yes,
<snip>

However On Friday 01 July 2005 13:02, Stephen Frost wrote:
> * Bruce Momjian (pgman(at)candle(dot)pha(dot)pa(dot)us) wrote:
> > Stupid question, but how do roles relate to our existing "groups"?
>
> Uhhh. There are no longer "groups", they've been replaced with roles
> (which can have members).
>

Was following this conversation up till now, because these two statement seem
to contradict each other. Do we really support groups still, are is CREATE
GROUP now syntactical sugar for some for of CREATE ROLE.

--
Robert Treat
Build A Brighter Lamp :: Linux Apache {middleware} PostgreSQL

Robert Treat <xzilla(at)users(dot)sourceforge(dot)net> writes:
> Was following this conversation up till now, because these two statement seem
> to contradict each other. Do we really support groups still, are is CREATE
> GROUP now syntactical sugar for some for of CREATE ROLE.

CREATE GROUP and CREATE USER are both now syntactic sugar for CREATE
ROLE.

regards, tom lane

* Robert Treat (xzilla(at)users(dot)sourceforge(dot)net) wrote:
> On Friday 01 July 2005 13:07, Stephen Frost wrote:
> However On Friday 01 July 2005 13:02, Stephen Frost wrote:
> > * Bruce Momjian (pgman(at)candle(dot)pha(dot)pa(dot)us) wrote:
> > > Stupid question, but how do roles relate to our existing "groups"?
> >
> > Uhhh. There are no longer "groups", they've been replaced with roles
> > (which can have members).
> >
>
> Was following this conversation up till now, because these two statement seem
> to contradict each other. Do we really support groups still, are is CREATE
> GROUP now syntactical sugar for some for of CREATE ROLE.

CREATE GROUP just does a CREATE ROLE now, yeah. You can check gram.y
for the details if you'd like. We do still support \du and \dg
(pg_users and pg_groups respectively, iirc) for backwards compat. and to
help folks get used to the new stuff.

Thanks,

Stephen

Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr> writes:
>> Right, this can be done now.

> There is the namespace collision issue, and although I might grant a
> student the privilege to create simple roles, I would not allow them to
> create new users for a basic practice;-)

Why not? With the setup Stephen suggests, they could create only new
users that could only get into their own database (since they'd not be
able to grant connect rights to other databases).

We probably need to think a bit harder about the meaning of CREATEROLE
though. Right now it gives free license not only to create roles but
to alter any property of existing roles. This seems appropriate if you
think of it as a "safer form of superuser", which is how I was thinking
of it. It would be too powerful for Fabien's situation though.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Fabien COELHO <coelho(at)cri(dot)ensmp(dot)fr> writes:
> >> Right, this can be done now.
> > There is the namespace collision issue, and although I might grant a
> > student the privilege to create simple roles, I would not allow them to
> > create new users for a basic practice;-)
>
> Why not? With the setup Stephen suggests, they could create only new
> users that could only get into their own database (since they'd not be
> able to grant connect rights to other databases).

I'm curious why not too... One issue I can think of is that
alter role ... rename, etc could be a problem.

> We probably need to think a bit harder about the meaning of CREATEROLE
> though. Right now it gives free license not only to create roles but
> to alter any property of existing roles. This seems appropriate if you
> think of it as a "safer form of superuser", which is how I was thinking
> of it. It would be too powerful for Fabien's situation though.

Well, what about it makes it 'too powerful'? I think that's pretty much
the same question you're asking Fabien above. I agree that only certain
properties should probably be modifiable though, one inparticular that
comes to mind is 'LOGIN'; I can see why you might want to allow only roles
which can't log in to be creatable by a given role.

It strikes me that it'd make some sense to have independent grant
control over each of the fields in pg_authid. It would also make sense
to limit the power of alter role to certain roles based on who they were
created by (superuser vs. createrole). It seems we probably need at
least an association to either catalog or creator for each role which
could then be used to limit alter role commands. catalog probably makes
more sense in the long run, creator would be easier in the short-term.

Thanks,

Stephen

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > Tom, if you're watching, are you working on this? I can probably spend
> > some time today on it, if that'd be helpful.
>
> I am not; I was hoping you'd deal with SET ROLE. Is it really much
> different from SET SESSION AUTHORIZATION?

Here's what I've got done so far on SET ROLE. I wasn't able to get as
much done as I'd hoped to. This is mostly just to get something posted
before the end of today in case some might think it's more of a feature
than a bug needing to be fixed (which is what I'd consider it,
personally). I'll try and work on it some this weekend, but in the US
it's a holiday weekend and I'm pretty busy. :/

Thanks,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> * Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
>> We probably need to think a bit harder about the meaning of CREATEROLE
>> though. Right now it gives free license not only to create roles but
>> to alter any property of existing roles. This seems appropriate if you
>> think of it as a "safer form of superuser", which is how I was thinking
>> of it. It would be too powerful for Fabien's situation though.

> Well, what about it makes it 'too powerful'?

The fact that you could alter roles that (in some sense) don't belong to
you. In particular you could allow yourself into other people's
databases, if you can alter the roles that they are using to define who
can connect to those databases. Or cause denials of service by revoking
other people's memberships in those roles.

On the other hand, CREATEROLE as-is does exactly what it was intended
to do, namely let the DBA do all normal admin maintenance of users/groups
without taking the risks involved in doing stuff as superuser. If we
restrict it more, then we'll be back to the situation where there are
routine manual admin tasks that require superuserdom. So on reflection
I don't think we should restrict it. If we need a more restrictive
feature, then we need a different feature.

I'm of the opinion that it's too late to do much about Fabien's use-case
in this devel cycle. We could possibly have designed something rational
if it had been brought up earlier in the discussion of roles (which I
remind you has been going on for months). But now it's too late to do
anything that wouldn't be a kluge, and probably a kluge we'd regret
later.

Possibly for 8.2 we could invent a notion of roles having owners.
Offhand I don't see any harm in letting non-CREATEROLE users create
non-login roles, and manipulate the membership of roles they have
created (or that have been assigned to them by a superuser). On the
other hand, it could be that the WITH ADMIN OPTION feature is already
sufficient for this. This really needs some thought ...

regards, tom lane

Dear Tom,

> We probably need to think a bit harder about the meaning of CREATEROLE
> though. Right now it gives free license not only to create roles but
> to alter any property of existing roles. This seems appropriate if you
> think of it as a "safer form of superuser", which is how I was thinking
> of it. It would be too powerful for Fabien's situation though.

Yes.

ISTM it would be a good thing to separate

- sysadmin issues (create a user that can connect - pg_hba.conf) and

- dbadmin issues (manage permission on data in a catalog).

The first item is the current pg superuser. I envision the second item as
being a property of the database (catalog) OWNER, which can do whatever
he/she wants at home, but should not bother others. So there is a need for
a limited 'createrole' ability (simple roles that cannot be users (?), and
avoid name space collision).

This is better for teaching, but also for any service provider which would
host many users and databases, and would like to transfer them.

If you want to delegate data access privileges to someone, it is better
to have that restricted to the catalog to avoid name-space collision.

The same issue arises when moving a database from one cluster to another
where some role names may already be used. That is the kind of think I
also need to do as a dbadmin.

If one take a conceptual perspective, database access privilege management
is a catalog-related task, so it seems better to have that in the catalog.

So I think all points that roles are better (more useful) per-catalog
than per-cluster.

If you just want to claim that 'pg has roles', you can do the marketing
with per-cluster roles;-) If you want them more useful, you need them
per-catalog.

--
Fabien.

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> Attached please find a patch to change how the permissions checking
> for alter-owner is done. With roles there can be more than one
> 'owner' of an object and therefore it becomes sensible to allow
> specific cases of ownership change for non-superusers.

Applied with minor revisions. The patch as submitted suffered a certain
amount of copy-and-paste-itis (eg, trying to use pg_type_ownercheck on
an opclass), and I really disliked using ACLCHECK_NOT_OWNER as the way
to report "you can't assign ownership to that role because you are not
a member of it". So I made a separate error message for that case.

regards, tom lane

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > Attached please find a patch to change how the permissions checking
> > for alter-owner is done. With roles there can be more than one
> > 'owner' of an object and therefore it becomes sensible to allow
> > specific cases of ownership change for non-superusers.
>
> Applied with minor revisions. The patch as submitted suffered a certain
> amount of copy-and-paste-itis (eg, trying to use pg_type_ownercheck on
> an opclass), and I really disliked using ACLCHECK_NOT_OWNER as the way
> to report "you can't assign ownership to that role because you are not
> a member of it". So I made a separate error message for that case.

Great, thanks! Sorry about the copy-and-paste-itis... Must have been a
case I wasn't sure about. The different error message makes perfect
sense. I see you also did the superuser-in-every-role change that I had
included, thanks.

When writing this patch it occurred to me that we nuke our
member-of-role cache for one-off lookups on occation. I don't
particularly like that, especially when we *know* it's a one-off lookup,
so I was considering adding a function for the one-off lookup case but
I couldn't come up with a way to avoid a fair bit of mostly-the-same
code as the current cache-regen code, without making the cache-regen
code alot slower which would negate the point.

Just some thoughts.

Thanks again,

Stephen

Stephen Frost <sfrost(at)snowman(dot)net> writes:
> When writing this patch it occurred to me that we nuke our
> member-of-role cache for one-off lookups on occation. I don't
> particularly like that, especially when we *know* it's a one-off lookup,

Yeah. What I had been thinking about is that maybe it shouldn't be just
a one-element cache. At the moment though we have no performance data
to justify complicating matters (heck, not even any to prove we need a
cache at all...) It'd be smart to try to profile some realistic cases
before spending any time coding.

regards, tom lane