Apparent timing problem with socket communications

["James H H Lampert" <jamesl@xxxxxxxxxxx>]

This really did get absolutely no response last week, on the Java list
(after what happened with my Unicode query before, I double-checked). And
once again, I'm not at liberty to be any more specific than I already
have. And I'm stumped on it, and unable to provide anything but a
palliative.

Fellow programmers:

I'm afraid I can't get into specifics here, but I've got a weird timing
problem with a Java client talking to an AS/400-native client (ILE C and
ILE RPG) through a socket, in an established connection. Under easily
duplicated conditions, the client sends a request to the server, then
fails to receive the reply sent by the server. Yet if I put diagnostics (a
series of printf statements in the C portion) into the part of the server
that processes the request and sends the reply, or even if I just have the
server wait for a fraction of a second, it works fine. In our own
installation. Yet it still fails at the customer site.

About the only specifics I can get into here are that the Java client is
reading the reply from a DataInputStream on a BufferedInputStream on the
stream returned from a getInputStream() on the socket.

It's my understanding that the BufferedInputStream layer should keep
anything from getting lost if the client isn't ready to listen when the
reply comes in. And yet that appears to be what's somehow happening. Any
suggestions on what could be going wrong?

(And I've already shared all the specifics I dare, so please don't ask for
more.)

--
James H. H. Lampert