Re: Run CU-SeeMe Client and Reflector Simultaneously

Brian Godette (bgodette@idcomm.com)
Thu, 19 Mar 1998 20:14:31 -0700


At 07:43 PM 3/19/98 BST, you wrote:
>** Reply to note from CU-SEEME-L@cornell.edu Thu, 19 Mar 1998 02:47:18 EST
>
>>
>> On Mon, 16 Mar 1998, Brian Godette wrote:
>> > Java-CU should really be called Java-NV since that is what it really is.
>> > It's simply the CU-SeeMe B&W codec handling and RTP communication
routines
>> > from NV 3.3 rewritten in Java. Since it uses seperate ports on the
client &
>> > server it can be run on the same system as the reflector.
>>
>> The German guy that wrote JavaCU provides the source. Is there any way for
>> JavaCU to be written to NOT require NV? I talked a bit with Andrew Clarke
>> about this awhile back. The docs to JavaCU mention something about there
>> being a 15kbps limit with Java applets. I'm not sure if that's true or
>> not. If it's not, a Java port of CU (though it would have to be receive
>> only I imagine) would be quite interesting to see.
>
>However, a 'true' CU client cannot be on the same ip address as the
reflector, unless the
>server it was on was multi-homed and the ref and client were bound to
different interfaces
>(possible in Java 1.1 and above).

Ummmm, sorta :) The current ERef code you have will accept source ports
other than 7648 (within reason, doesn't go below 1024) from clients and
will send it back to that port. This of course is in preperation for a few
other projects floating around out there <wink> <wink>.

>On a related note (and for a change this *isn't* a slight against White
Pine) this whole
>cross-platform issue is interestng. Brian's ref to my sure knowledge is
running on three
>different operating systems, Linux/Unix, Win32 and OS/2, because it's
written in very standard
>ANSI-C. As an example, porting the ref to OS/2 took me the best part of 2
hours, which is
>entirely down to Brian's clean code, I only added the #ifdefs for the OS/2
specific parts
>which don't number highly at all. I'd be thrilled if WhitePine created
some Java-based CU

I seem to remember it being nothing more than than the OS/2 specific TCP/IP
startup code (ala WSAStartup under win32) and OS/2 specific close routine
for sockets (soclose() under OS/2, closesocket() under win32, and close()
under Unix), everything else remained the same. And so far the only "issue"
the seems to pop up is sys_errlist being already defined under certain
flavors of Unix (MkLinux and one or two others). Even compiles and runs
cleaning on DEC Alpha's (64 bit Code) running OSF/1 or Digital Unix.