Hello Michelle, > Fatal transport error: Address already in use: NET_Bind > java.net.BindException: Address already in use: NET_Bind > at java.net.PlainSocketImpl.bind(PlainSocketImpl.java:398) > > I also wrote the code use apache client directly and it also causes > the same exception. > > this made me think this may be an apache bug.
Socket opening is not under control of HttpClient. HttpClient calls a ProtocolSocketFactory, which calls a standard Java API to open a socket. Apparently, HttpURLConnection does something differently. If it is known what the difference is, there is a chance of adding a workaround to HttpClient, or at least to document how to work around the problem. Have you tried to rip out all the HTTP communication and just perform the socket opening and closing alone? Maybe with sending some hard-coded request. I'm sure you'll be able to reproduce this problem without any HttpClient code being executed. > The wield thing is we > can only recreate this problem in some of vista machines, but not > all vista machines. And it works fine on windows XP. Could this be > an apache bug? No, because the HttpClient code seems to work fine on all other platforms. At most, it is an incompatibility of the HttpClient code with Vista. In that case, it is up for debate whether Vista/JVM or HttpClient should be changed. But changing HttpClient - if we knew what needs to be done - means changing the behavior on *all* platforms, which is a higher risk. > if so, what is the process to open an apache bug? For HttpClient, issues are tracked in JIRA: http://jakarta.apache.org/commons/httpclient/issue-tracking.html You can open an issue for tracking purposes. But you should understand that the developer community of HttpClient is rather small, and probably does not have the resources to recreate this problem on Vista. In particular since you report that it occurs only on *some* of your Vista machines. FWIW, I think your best chance to get this resolved is to re-create the problem without HttpClient, just by using plain old sockets. Then you can report it to Hursely. Btw, are these problems with the full SUN/IBM JVM, or with J9? > Also, is there anyway to tune the system settings instead of > changing the code as you recommended below to solve this problem? That can only be answered by someone with access to a Vista system where the problem can be reproduced. cheers, Roland
smime.p7s
Description: S/MIME Cryptographic Signature
