[EMAIL PROTECTED] wrote: > I thought many other things about DIRMINA-42: > -> A single IoAcceptor can manage (and bind/unbind) many > localSocketAddress, so when we unbind only one of them we have to > disconnect only the IoSessions correlated to that SocketAddress > (DIRMINA-93). Because of this is good to mantain a IoSessionMap for each > binded service in IoAcceptor.
Yes, the IoAcceptors whould definitely need to maintain a Map of IoSessions with the local address (the port bound to) as key. I'm not sure if this Map would have to be exposed to others though. Maybe it's better to have a method with the signature Collection getSessions(SocketAddress) in IoAcceptor. One problem is what to do with IoConnectors. They are IoSessionManagers as well. In that case it's not as obvious what should be the key in the Map. I'm not sure if we need IoConnectors to maintain a list of sessions. Maybe we could skip that for know? > -> The choice between hashMap or simple collection is interesting, but > many codes can be developed in the same manner, so i begun to write down > some interfaces and some patches to mina9.0 code, patching the tcp > packages especially. Great! > May I send that codes on jira? I read something about the Apache > License, have I to modifie the code, apply some introduction to each class? Yes of course, please attach the code to the JIRA issue. The coding conventions used by MINA are described on http://directory.apache.org/community/process/code-standards.html I'm not sure if it's 100% up to date. I also have coding style and formatting files which you can use with Eclipse. I will send them to you off list. /Niklas
