X-Posted IBM-MAIN and IBMTCP-L. A (z/OS, if it matters) program executes the following pseudo-code on a blocking TCP/IP socket:
len = send(socket, msgA, lenA, flags); if ( len != lenA ) an error has occurred len = send(socket, msgB, lenB, flags); if ( len != lenB ) an error has occurred Let's assume that the connection has failed a little *before* the send of *msgA* -- perhaps the session partner was re-booted or powered off, or a router was reset. Questions: 1. Based on observation I have concluded that the send of msgA *never* errors (len != lenA), but that sometimes or perhaps always the send of msgB errors. Is this consistent with others' observations and/or knowledge? Is this documented somewhere? 2. If so, will the send of msgB *always* error, or is it possible that the send of msgB will also appear to succeed and the program receive no error indication until sending one or more additional messages beyond msgB? 3. If so, once the program does get a send error, is it possible for the program to determine how many messages were not delivered (absent an application level "ack" or "restart point" protocol)? Charles ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN