Santiago Gala wrote:

 > Paul Spencer wrote:
 >
 > (I snip most. I agree with most of it)
 >
 >>
 >>
 >> Missing group = all user have access based on roles.
 >
 >
 > As turbine roles are defined in a triple group-role-user, this
 > should  better be rephrased as "all users have access based on
 > roles in global  group". I'm not picky, this is important. This
 > misunderstanding is the  root of most problems we have experienced.
 >  A user has *no* role unless  you pick up a group.
 >

 >>
 >> Missing roles = all user have "default access"
 >
 >
 > This could be the users have the permissions returned by 
data.getACL().hasPermission("xxx"),
 >  but, again, this is implemented in  turbine as return
 > hasPermission(permission,  TurbineSecurity.getGlobalGroup());, so
 > it means "access according to  permissions in the global group" The
 >  precise meaning of a role constraint is "data.getACL().hasRole(
 > role  ) && data.getACL().hasPermission( "xxx" ) returns true",
 > where xxx is  the action to be performed. It restricts the
 > permission further, since I  could have view permission, but not
 > clerk role, for instance.
 >

Since the user creation and group/role assignment is managed through 
Jetspeed, we can enforce a user configuration the does not utilize any 
group or role defaults.  In addition, we should document "no group" and 
"no role" as unsupported.

 >>
 >>
 >> We also need to define a standard set of permissions: view       -
 >>  Allows portlet content to be viewed and added to pane customize
 >>  - Allows portlet to be customized
 >
 >
 > This can/should be required at the portletset level

This should be allowed.  Granted the portlet type can enable of disable 
a permission, like customize.  I think the ability to say "this portlet 
can be customized" in the registry entry is valuable.  Doing this by 
adding permissions to the <portlet-entry> is one way to do this. Another 
is with parameters that are utilized by the allowCustomize type methods.

 >
 >>
 >> minimize   - allows portlet to be minimized
 >>
 >> maximize   - allows portlet to be maximized remove     - allows
 >> portlet to be removed from pane
 >>
 >> move       - allows portlet to be moved in pane
 >
 >
 > This one looks very difficult to implement, since in a two column 
layout,
 >  moving one portlet shuffles the rest (effectively moving them).  Ideas?
 >

Portlets can not move past a non-moveable portlet.  This still leave a 
few holes. Documented Known Problems/restrictions :)
  o If the non-moveable portlet is not in the first row, then
    preceding portlet can be deleted. Thus moving the non-moveable
    portlet up.
  o If the non-moveable portlet is not in the first row, then
    the portlet can be moved down by a portlet that is moved to
    the first row from an adjacent column.

 >
 >>
 >>
 >>
 >> One other area the is related, it the ability to have common portlet(s)
 >>  the are maintained in one place for a group of users.  As  an
 >> example corporate E-mail and corporate News should be on all
 >> pages  and in the left column.  (I am not ask this be addressed
 >> now, just be  aware of this desired functionality while designing
 >>  the security changes)
 >
 >
 > This again could be implemented with relative ease at the portletset 
  level, except that a multi-column layout is *not* a portletset. I mean
 > something like the customiser not allowing to modify a given part of the
 > PSML. If a Document is composed of a (role=admin)set which includes
 > several (unrestricted)portletsets, the user could modify only the inner
 > ones, but should respect the structure of the outer one. We will work
 > security in the PSML structure/customiser later on, when we have the
 > basics sorted out.
 >

I expect the customizer will need to support nesting portlet sets.  As a 
enable having a "row" that contains a "column" and a "multi-column" 
portletset. But this is outside the scope of "Security changes"

 >
 >
 >

Paul Spencer


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to