[keycloak-dev] Expected behaviour for rememberMe?
Stian Thorgersen
sthorger at redhat.com
Mon Apr 4 04:50:47 EDT 2016
On 4 April 2016 at 10:44, Stian Thorgersen <sthorger at redhat.com> wrote:
>
> On 4 April 2016 at 09:31, Marek Posolda <mposolda at redhat.com> wrote:
>
>> Seems there are 2 things here:
>>
>> 1) Username "hint" provided by KEYCLOAK_REMEMBERME cookie. IMO this
>> cookie should be deleted only when:
>> - User explicitly clicked on logout and maually logout himself
>> - User click on "Login" button on login screen without the rememberme
>> checkbox checked
>>
>> IMO it shouldn't be deleted when SSO cookie is expired, which is current
>> behaviour and should be changed IMO. In other words, I expect the scenario
>> working like:
>> - User logged with "rememberMe" checkbox on
>> - User closed the browser
>> - After a month, user returned back to the application. His SSO session
>> is expired, but KEYCLOAK_REMEMBERME cookie won't be deleted, so on login
>> screen he will see the prefilled username and rememberMe checkbox switched
>> to "on"
>>
>
> Create a JIRA to request remember me cookie to not be removed. However, we
> need some way of configuring expiration of the cookie. This would be for
> 2.x.
>
>
>>
>>
>> 2) Persistent KEYCLOAK_IDENTITY cookie when rememberMe is switched to on.
>> I can't see how it can work when session is expired as it relies on session
>> in the cookie value. On the other hand, rememberMe shouldn't rely on "SSO
>> Session idle timeout" IMO. SSO IDle timeout is only 30 minutes by default.
>> So current behaviour is, that when user closes his browser, he needs to
>> open in again and being re-authenticated only when he do within 30 minutes,
>> which is bit of pointless IMO.
>>
>> I would suggest to change the behaviour like this:
>> - When userSession is marked as rememberMe, then cleaner thread will take
>> into account just "SSO Max Lifespan" timeout, but not SSO Idle timeout
>> - During verification of SSO cookie re-authentication and when session is
>> rememberMe, we will take into account just SSO Max Lifespan of session, but
>> not SSO Idle timeout
>> Refreshing of tokens will still take SSO Idle timeout just like now.
>>
>> If we not change the behaviour like this, we should at least update
>> "RememberMe" docs and tooltip to make it more clear what the behaviour
>> would be in various cases.
>> WDYT?
>>
>
> We've already discussed this and there's a JIRA requesting it (
> https://issues.jboss.org/browse/KEYCLOAK-1267). The default behavior
> should be that SSO Idle timeout is taken into account, but there should be
> an realm option to ignore it and only rely on SSO Max lifespan. This is
> also for 2.x.
>
Actually, thinking about this some more IMO we should either reject
KEYCLOAK-1267 or add separate idle/max configuration for remember me, not
just ignore. Having user sessions that doesn't take SSO Idle into account
would potentially result in a large number of unused user sessions left in
the system. Especially if SSO Max is large. It could be users clicked it by
mistake in incognito mode, they manually cleared cookies, they re-installed
the machine, etc.
>
>
>>
>> Marek
>>
>>
>> On 31/03/16 16:26, Libor Krzyzanek wrote:
>>
>> I read docs today
>> http://keycloak.github.io/docs/userguide/keycloak-server/html/timeouts.html#d4e2630
>> and my understanding is that user should keep logged in after either
>> browser restart or session expiration.
>> My tests shows that after session expiration (set to 1 min) I have to log
>> in again.
>>
>> Thanks,
>>
>> Libor Krzyžanek
>> Principal Software Engineer
>> Red Hat Developers | Engineering
>>
>> On Mar 31, 2016, at 3:00 PM, Marek Posolda <mposolda at redhat.com> wrote:
>>
>> Followup on the issue by Libor [1] . I can confirm to see the same
>> behaviour in the OOTB Keycloak, like Libor described in the JIRA. In
>> other words, when you refresh account page (
>> http://localhost:8080/auth/realms/myrealm/account ) but the UserSession
>> referenced from KEYCLOAK_IDENTITY cookie is expired, then all cookies
>> including KEYCLOAK_REMEMBERME are expired too.
>>
>> IMO RememberMe cookie shouldn't be expired when session is expired.
>> We're using the rememberMe cookie as hint for username on the login
>> page. So even if user returns to page after a month, I am not seeing
>> anything bad that rememberMe cookie is still valid and user will see
>> "hint" with his username on login page and rememberMe checkbox checked
>> even if session was expired already for a long time. IMO the only
>> situation when we should expire KEYCLOAK_REMEMBERME cookie is, when user
>> unchecks the "Remember me" checkbox on login page.
>>
>> [1] https://issues.jboss.org/browse/ORG-2956
>>
>> Marek
>> _______________________________________________
>> keycloak-dev mailing list
>> keycloak-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/keycloak-dev
>>
>>
>>
>>
>> _______________________________________________
>> keycloak-dev mailing list
>> keycloak-dev at lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/keycloak-dev
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/keycloak-dev/attachments/20160404/daece776/attachment.html
More information about the keycloak-dev
mailing list