2008/8/13 Murat OZDEMiR <[EMAIL PROTECTED]>: > Hi Aidan, > >>> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf > Of Aidan Skinner >>> Sent: Monday, August 11, 2008 12:06 AM >>> MINA 2.0.0 is still some way away from being stable AFAICT, there are >>> also a couple of problems with mina 1.0.1 which prevent use from using >>> it. > > I've took a look at > http://issues.apache.org/jira/browse/DIRMINA > in Versions Page i found the issues for MINA 1.0.1; > http://issues.apache.org/jira/browse/DIRMINA/fixforversion/12312080 > > DIRMINA-307 FIXED ReadThrottleFilterBuilder does not resume reading. > DIRMINA-283 FIXED Sessions are not recycled in IPv6 environment when > bind interface is not specified. > DIRMINA-275 FIXED Exception in thread "SocketAcceptor-1" > java.nio.channels.CancelledKeyException > DIRMINA-273 FIXED Session created with non-existent service > DIRMINA-266 FIXED SocketAcceptor.bind() requires address != null and > port != 0... why? > Also can get to > ServerSocketChannel.socket().getLocalSocketAddress() > DIRMINA-235 INVALID Reliable hang of DS during query > DIRMINA-284 FIXED Filter.sessionCreated/Filter.sessionOpened called > AFTER Filter.filterWrite... > DIRMINA-285 FIXED Incorrect calculation of IoSession.writtenMessages() > > DIRMINA-308 FIXED NullPointerException in SocketIoProcessor under > heavy load > DIRMINA-253 FIXED Update FAQ > DIRMINA-313 FIXED LoggingFilter logs exceptionCaught events on error > level but uses isInfoEnabled in the if-statement > DIRMINA-277 FIXED Various tests block with JDK5 concurrent used in > place of backport > DIRMINA-288 FIXED Add SessionLog.info( session, "CREATED" ); to > LoggingFilter.sessionCreated method > > I understand why don't you want to use MINA2 but why do you abstain to use > 1.0.1 or newer for example 1.1.8? > > Murat
Murat, The problem is that all versions beyond 1.0.1 are incompatible with Qpid due to a 'Performance' change that was introduced in all subsequent versions. In the pre 1.0 days we worked with mina to address an incompatibility in their code that prevented us slicing buffers returned from mina. https://issues.apache.org/jira/browse/DIRMINA-201 This change however was reverted by a change that made it in to all subsequent releases. This change IMO defaults the mina code to save memory in a very small use case. https://issues.apache.org/jira/browse/DIRMINA-328 https://issues.apache.org/jira/browse/DIRMINA-346 As commented on DIRMINA-346 this reintroduces 201. I don't believe we pointed this out to the team at the time as Mina was performing adequately and we had no need to upgrade. I believe that I have mentioned it on the mina list but we have no pressing need to upgrade our mina I think we are more likely to migrate away from mina as there is quite an overhead associated with it that we don't see with our own IO. As always patches are welcome I'm sure the biggest issues after getting the above bug addressed will be any API changes and as Aidan has pointed out Mina 2.0 is still a beta and we can only release from Apache based on GA releases. If you have a patch available that shows 2.0 to be much faster that 1.0 then that would be most welcome. hth Martin -- Martin Ritchie
