[security-dev] Group clarification

[Shane Bryzak]

On 07/02/13 11:21, Bolesław Dawidowicz wrote:
> We had some chat on IRC and it turned out there is also currently no
> method to obtain the "path ID" of the group. I wondered why it was dropped.

It was dropped because we refactored the identity model so that all 
identity classes had a unique id (UID) value. We should probably add a 
different method for Group that returns the "fully qualified" name, but 
it can't be called getId() now.
>
> For me if we allow only one parent group it makes the model natural tree
> hierarchy anyway. Then if you allow /a/b/c/a/b (5 different groups) it
> is handy to be able to obtain such path in a single call. However I
> think it should be really stored on a group level. Calculating it on
> demand can be performance concern. Deep trees are not that often design.
> In most LDAP schemas I witnessed deepest was around 3-4 levels. However
> if it happens then such calculation will result in N calls to DB/LDAP.
>
> Something like this would be handy IMO:
>
> identityManager.getGroupById("/a/b/c/a/b").getId().equals("/a/b/c/a/b");
>
> it can be path or key instead of id if it fits current design better.
>
> I also wonder about groups without hierarchy (no parent) scenario. If
> there should be null parent or some special root entity.
>
> On 02/07/2013 05:03 PM, Shane Bryzak wrote:
>> We should already support it.  If it doesn't work, please raise an issue
>> in jira (PLINK) and assign to me.
>>
>> On 07/02/13 07:02, Marek Posolda wrote:
>>> Hello,
>>>
>>> One of the current requirements in GateIn is possibility to have groups
>>> with same name and with different parents. For example: I can have
>>> groups "/qa/management" and "/dev/management"
>>>
>>> In other words, I have two groups called "management" but both are in
>>> different parts of group tree, because first one has parent group "qa"
>>> and second has parent group "dev". Currently Picketlink IDM 3 doesn't
>>> support it (it always throws exception when it recognize that group with
>>> same name already exists). Also I am seeing that concept of GroupID
>>> (path to group from root group - something like "/qa/management") and
>>> group key has been removed as well even if it was supported in IDM 3.x
>>> couple of weeks before.
>>>
>>> Also for read usecase, there are two methods in IdentityManager to find
>>> groups:
>>>
>>>         Group getGroup(String groupId);
>>>
>>>         Group getGroup(String groupName, Group parent);
>>>
>>> I think that first one has been designed to find group with argument as
>>> groupId, so usage could looks like:
>>>
>>> Group qaManagersGroup = identityManager.getGroup("/qa/management");
>>>
>>> Second one has been designed with usage of plain group names like:
>>>
>>> Group qaGroup = identityManager.getGroup("qa", null);
>>> Group qaManagersGroup = identityManager.getGroup("management", qaGroup);
>>>
>>>
>>> Problem is that currently we are always using first one with groupName
>>> as an argument (not groupId), so it obviously can't work correctly if we
>>> have two groups with same name "management" because it's unclear which
>>> one should be result of finding...:-\
>>>
>>>
>>> Any ideas to address this? My current proposal is:
>>>
>>> - Return concept of groupId, which will return the path like
>>> "/qa/management". So usage could be like:
>>> Group qaGroup = new SimpleGroup("qa");
>>> Group qaManagementGroup = new SimpleGroup("management", qaGroup);
>>> assertEquals("management", qaManagementGroup.getName());
>>> assertEquals("/qa/management", qaManagement);
>>>
>>> - Either
>>> -- fix all existing usages of identityManager.getGroup(String groupId),
>>> so it really expects groupId as argument (not groupName):
>>>
>>> -- or introduce new method on IdentityManager (and IdentityStore) like:
>>>
>>> Group getGroupByGroupId(String groupId);
>>>
>>> It's possible that some identityStore implementations doesn't support
>>> groups with same name (For example current LDAPIdentityStore can't
>>> support it because there is only one DN for access all groups, but we
>>> discussed with Pedro that this is planned to address later)
>>>
>>> Any thoughts?
>>> Marek
>>> _______________________________________________
>>> security-dev mailing list
>>> security-dev at lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/security-dev
>> _______________________________________________
>> security-dev mailing list
>> security-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/security-dev
>>
> _______________________________________________
> security-dev mailing list
> security-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/security-dev