From: Alan K. <jyt...@xh...> - 2008-03-06 14:25:08
|
[Bob] > I've opened a problem on these. It is issue 1005. Thanks Bob, I've assigned that to myself. [Bob] > By the way, I have been running some functional type tests with select, > asyncore, asynchat, and some other more 'exoteric' Python networking modules > and for Networking code that is really in pre-Beta it is amazingly stable. I'm glad to hear that your experience has been positive. [Bob] > I haven't done much yet on performance, but that is coming next. I am > hopeful for good performance because since Jython networking (unless I am > misunderstanding) will now always use NIO channels. Is that true? Jython networking *mostly* uses java.nio, but there are some situations where java.net *must* be used, e.g. timeout support on UDP sockets, for example; the java.nio methods simply do not honour timeouts on UDP sockets. This is why the jython socket implementation is so messy, having to use java.nio in some circumstances and java.net in others. > If so, I > will run some comparisons with Java io. It will be interesting to compare > Jython NIO performance with Java net performance. Indeed, it would be most interesting; please do share the results with us. [Bob] > I tried to assign the problem to Alan, but it doesn't look like the > assignment was successful. I don't know if I am supposed to have authority > to assign problems or not since I am not a committer. No worries, I've assigned it to myself already. Thanks, Alan. |