Would this new solution be compatible with URL rewriting? (No cookies 
being used)

[EMAIL PROTECTED] wrote:
> On Mon, 8 Jul 2002, Denis Benoit wrote:
> 
> 
>>I think it would be difficult, since JSESSIONID is distinct for each
>>webapp on a Tomcat, only JSESSIONIDSSO (if the SingleSignon valve
>>is activated) is common to all webapps.
>>
>>I'll try to think of something, but if you think of something first,
>>let me know :)
> 
> 
> Well, my thinking is that in order to have 'single signon' you need 
> a way to have a single cookie ( or path param if cookies are disabled )
> across all webapps. Whatever mean to get that as JSESSIONIDSSO, 
> it can be used for JSESSIONID as well.
> 
> So I would add a hook into the session id generator - and have
> the single signon use the hook to push session ids.
> 
> If we want to have distinct sessions in each webapp - the session
> id would consist of the 'common' part and a per-webapp part.
> 
> In general, my view of single signon is that each app must 
> redirect to an auth application ( similar with kerberos for example)
> and use the certificate as session id for all webapps.
> 
> Costin
> 
> 
> 
>>On Mon, 8 Jul 2002 [EMAIL PROTECTED] wrote:
>>
>>
>>>+1
>>>
>>>But before doing that - would it be possible to replace JSESSIONIDSSO 
>>>with a mechanism relying only on JSESSIONID ? 
>>>
>>>Even if we patch mod_jk, there are other load balancing solutions
>>>( hardware, etc ) - it would be much simpler if from 'outside'
>>>we would only use the standard JSESSIONID cookie / path param.
>>>
>>>Costin
>>>
>>>On Thu, 4 Jul 2002, Denis Benoit wrote:
>>>
>>>
>>>>Hi,
>>>>
>>>>With the current code (TC 4.1.6), the single signon does not work with the
>>>>loadbalancer connector.
>>>>
>>>>If a user was logged in a given webapp, the loadbalancer looks at the
>>>>JSESSIONID cookie (or URL parameter) to dispatch the request properly to the
>>>>tomcat where the user was logged on.  But if the user hits another webapp,
>>>>the JSESSIONID is not present anymore and the dispatcher applies its
>>>>round-robin logic to dispatch the request to any tomcat.  It nullifies the
>>>>effect of the single signon.  There is two problem that prevent it to work.
>>>>
>>>>1. On the Tomcat side, the generateSessionId() method of
>>>>   org.apache.catalina.authenticator.AuthenticatorBase does not append
>>>>   the jvmRoute of the Engine if one is specified.  So when a user changes
>>>>   webapp, the web connector dispatcher does not have any information to
>>>>   properly route the request;
>>>>
>>>>2. The current loadbalancer code specifically look for the JSESSIONID cookie
>>>>   and does not look for a JSESSIONIDSSO cookie.
>>>>
>>>>I could provide a patch to org.apache.catalina.authenticator.AuthenticatorBase
>>>>to add the jvmRoute to the session id; in fact it is a copy of the code from
>>>>org.apache.catalina.session.ManagerBase.
>>>>
>>>>The change in:
>>>>
>>>>    ./jk/native/common/jk_lb_worker.c
>>>>    ./jk/native2/common/jk_requtil.c
>>>>
>>>>is also trivial, first the connector must look for the JSESSIONID cookie (or
>>>>param), and if not found it should look for the JSESSIONIDSSO cookie (or
>>>>param).  Then the same logic should be applied if either one is found.
>>>>
>>>>Comments?
>>>>



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

Reply via email to