Note: need to set env vars before using the .vcproj file. See readme.txt in usr/terri These are the test runs I tried: 1. TAO ORB to TAO ORB on Windows (1gig RAM machine). 30,000,000 bytes sent and received OK multiple times. 100,000,000 did not work however. 2. JacORB server and TAO ORB client on windows. 40,000,000 send and received OK. Also tried JacORB client and server up to 40,000,000 OK. 3. JacORB to JacORB on Mac OS (A powerbook with 1 gig and Java 1.4.2). Can send 30,000,000 back and forth but only when I start both JVMs with "-server". Using "-server" increases the default heap size from 64M to 128M. Otherwise I get an "out of memory" exception on 20,000,000 sequences. Looks like a JVM tuning issue, not a CORBA limitation. 4. TAO to TAO on Mac OS up to 40,000,000 OK. 5. omniORBPy worked up to 80,000,000 on Mac iff "-ORBgiopMaxMsgSize 80001000" arg was used. ==================================================== Changes ==================================================== previous notes by mth (before 1 found the ORBgiopMaxMsgSize arg in omniORB): omniORB seems to be a weak link. Tested on mac (10.3.8, 1 Gig RAM). connecting client to C++ server 2097067 was the limit. If the message was over that the client got an exception: omniORB.CORBA.MARSHAL: Minor: MARSHAL_MessageSizeExceedLimitOnClient, COMPLETED_NO exception More troubling was the behavior when the server was written in python: messages length <= 2097060 worked. 2097060 < message len < 2097067 client recieves exception: omniORB.CORBA.MARSHAL: Minor: MARSHAL_PassEndOfMessage, COMPLETED_NO. 2097068 < message len<= 2097075 client hangs and must be killed The server console shows the message: omniORB: Unknown exception caught by a server thread at ../../../../../src/lib/omniORB/orbcore/GIOP_S.cc: line 271 (debugging output indicates that this occurs before my python servant's methods are called). The server does NOT need to be relaunched in order to work again. messages over 2097076 generate an exception: omniORB.CORBA.MARSHAL: Minor: MARSHAL_MessageSizeExceedLimitOnClient, COMPLETED_NO. when the C++ client was used with the python server the limit seemed to be 2097060. anything over that cause the server to hang with the omniORB: Unknown exception caught by a server thread at ../../../../../src/lib/omniORB/orbcore/GIOP_S.cc: line 271 error.