Hi,

Thanks - unfortunately my management forbids shared state for now, but it will 
be re-considered in a few months when our architecture matures.
But a million thanks for your willingness to dig up & share this code, it's 
deeply appreciated.


As per your remark - please note https://issues.jasig.org/browse/CAS-742  
*could*help with sign out behind a load-balancer, provided CAS can access 
internal node addresses (not through the load-balancer). 
It boils down to allowing business applications to add parameters when 
redirecting to CAS login, then extending SimpleWebApplicationServiceImpl  to 
store those parameters.
For example: 
https://server/cas/login?service=myInternalAddress&redirectTo=myExternalAddress
Now:
If "SimpleWebApplicationServiceImpl" could record this extra data, it could be 
used so that "myInternalAddress" (=internal node address behind the LB) can be 
used for Service Ticket validation + logout notifications, while 
"myExternalAddress" can be used only when redirecting the user back to the 
application on successful login.

I did a very rudimentary POC which worked, but it's not mature enough for 
contribution:
1) It didn't consider proxy callbacks, SAML, Google Accounts and various other 
features (we have just plain HTTP requests)
2) It's hard to change the instantiation of SimpleWebApplicationServiceImpl: 
it's not instantiated only through the Factory method (crateServiceFrom) but 
also from some direct constructor invocations
3) One might also want to encrypt/obscure the "internalAddress" param, if they 
don't want to expose their internal architecture

In any case, thanks again for taking the time to reply, it's really appreciated 
:)





----- Original Message -----
From: Marvin Addison <marvin.addi...@gmail.com>
To: cas-user@lists.jasig.org
Cc: 
Sent: Monday, December 10, 2012 4:26 PM
Subject: Re: [cas-user] Single Sign Out - and load balancer

> But please note solutions were already outlined:
> https://issues.jasig.org/browse/CAS-742

That issue in no way addresses single sign-out matters when the client
is clustered/load balanced.

> http://comments.gmane.org/gmane.comp.java.jasig.cas.devel/1495

That is a good reference. The patch I offered took the strategy of
sharing state across all client nodes. I think rebroadcasting is more
work and has more potential points of failure.

I thought I'd lost the patch, but it's actually attached to
https://issues.jasig.org/browse/CASC-114. I was perhaps a little
bashful at the time and didn't push it, but if we want to revisit the
issue now I'm perfectly happy to reignite the discussion.

> 4) Bottom line, I felt it's not delayed due to lack of contributors.

Disagree. I implemented and tested a solution and there was lackluster
interest and it withered on the vine. It takes more than a patch to
bring a feature of this scope into the product. There's a place for
code review, testing, and folks that simply offer words of support.
I'm sure there's a role for you in that list somewhere.

M

-- 
You are currently subscribed to cas-user@lists.jasig.org as: solmy...@yahoo.com
To unsubscribe, change settings or access archives, see 
http://www.ja-sig.org/wiki/display/JSG/cas-user


-- 
You are currently subscribed to cas-user@lists.jasig.org as: 
arch...@mail-archive.com
To unsubscribe, change settings or access archives, see 
http://www.ja-sig.org/wiki/display/JSG/cas-user

Reply via email to