-
30. Re: 1.4.1.GA ready
timfox Nov 20, 2008 4:01 AM (in response to clebert.suconic)"clebert.suconic@jboss.com" wrote:
Anil... are you sure this won't affect performance?
We must do a before and after performance check before applying those changes. If it's detrimental to performance it's a no go and we'll have to find some other way. -
31. Re: 1.4.1.GA ready
flavia.rainone Nov 20, 2008 7:32 AM (in response to clebert.suconic)Tim, I'm going to attach a new aop jar to the issue Anil created:
https://jira.jboss.org/jira/browse/JBAOP-675
so you can test for performance. I'll post a message here when I'm done. -
32. Re: 1.4.1.GA ready
flavia.rainone Nov 20, 2008 7:44 AM (in response to clebert.suconic)Tim, I'm hoping that your tests will show no significant impact on performance. The security priviledged block is going to execute only the first time the joinpoint is reached:
http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4190861#4190861 -
33. Re: 1.4.1.GA ready
flavia.rainone Nov 20, 2008 8:46 AM (in response to clebert.suconic)"flavia.rainone@jboss.com" wrote:
Tim, I'm hoping that your tests will show no significant impact on performance. The security priviledged block is going to execute only the first time the joinpoint is reached
Actually I'm mistaken. It will execute only the first time the joinpoint is reached on a per instance basis, given your aspects have a PER_INSTANCE scope.
Now, the performance impact depends on how many instances you have running and on the structure of your code. The jar is attached:
https://jira.jboss.org/jira/browse/JBAOP-675
Let me know what is the outcome of your performance tests. -
34. Re: 1.4.1.GA ready
clebert.suconic Nov 20, 2008 10:40 AM (in response to clebert.suconic)"timfox" wrote:
"clebert.suconic@jboss.com" wrote:
Anil... are you sure this won't affect performance?
We must do a before and after performance check before applying those changes. If it's detrimental to performance it's a no go and we'll have to find some other way.
IMO all we need is to make sure we aways test for System.getSecurityManager() != null before creating the Priviledged block. Then we know that users who care about performance won't be affected, and users who care about security on the VM will assume some little impact.
It's aways about managing expectations, right? -
35. Re: 1.4.1.GA ready
anil.saldhana Nov 20, 2008 10:44 AM (in response to clebert.suconic)Why not just recompile the JDK code base to remove the sandbox? This way we get the optimal performance that you desire. ;)
It is not a question of performance here. It is a question of shipping your flagship software with a security issue underneath. :)
If you say, performance is a requirement for shipping your software, I will also not allow you to release until this security issue is fixed. ;) -
36. Re: 1.4.1.GA ready
timfox Nov 20, 2008 10:47 AM (in response to clebert.suconic)I never said they were mutually exclusive.
-
37. Re: 1.4.1.GA ready
flavia.rainone Nov 20, 2008 11:07 AM (in response to clebert.suconic)"clebert.suconic@jboss.com" wrote:
IMO all we need is to make sure we aways test for System.getSecurityManager() != null before creating the Priviledged block. Then we know that users who care about performance won't be affected, and users who care about security on the VM will assume some little impact.
That's fine. All privileged blocks in AOP are preceded by a System.getSecurityManager() != null check. -
38. Re: 1.4.1.GA ready
ron_sigal Nov 20, 2008 12:09 PM (in response to clebert.suconic)I don't know if this is relevant to JBossMessaging, but I just noticed that the latest version of jnpserver.jar has a bunch of JBoss specific permissions, e.g.,
permission org.jboss.naming.JndiPermission "detection", "createSubcontext"; permission org.jboss.naming.JndiPermission "detection", "listBindings"; permission org.jboss.naming.JndiPermission "detection", "lookup"; permission org.jboss.naming.JndiPermission "detection/*", "rebind"; permission org.jboss.naming.JndiPermission "detection/*", "unbind";
-Ron -
39. Re: 1.4.1.GA ready
anil.saldhana Nov 20, 2008 12:18 PM (in response to clebert.suconic)The JIRA issue for that is etched in memory - JBNAME-8
-
40. Re: 1.4.1.GA ready
ron_sigal Nov 20, 2008 12:26 PM (in response to clebert.suconic)Of course, I'm the last to know. :)
-
41. Re: 1.4.1.GA ready
clebert.suconic Nov 20, 2008 10:43 PM (in response to clebert.suconic)"anil.saldhana@jboss.com" wrote:
in the AS5 workspace, the thirdparty/jboss/messaging contains both "jboss-messaging.jar and jboss-messaging-client.jar"
but the client directory of the built AS ie. build/output/jboss-5.x/client directory contains jboss-messaging.jar rather than jboss-messaging-client.jar. Is this intended?
Hoping Scott/Dimitris would see this...
Where this should be fixed on the build?
that jboss-messaging.jar on /client really is the server library. -
42. Re: 1.4.1.GA ready
starksm64 Nov 20, 2008 10:56 PM (in response to clebert.suconic)We can change it, but would need to verify that there are no assumptions about classes in the server jar being needed by the tck tests, etc.
-
43. Re: 1.4.1.GA ready
anil.saldhana Nov 21, 2008 2:25 AM (in response to clebert.suconic)Its quite late in the game for AS5GA. But it needs to be fixed if there is a jboss-messaging-client.jar available.
Ping Dimitris/Rajesh and get this sorted out. -
44. Re: 1.4.1.GA ready
clebert.suconic Nov 21, 2008 10:09 AM (in response to clebert.suconic)Dimitris: ping :-)
https://jira.jboss.org/jira/browse/JBAS-6231