On Wednesday, August 27, 2003, at 02:50 PM, Dar Scott wrote:
I have a small sockets stack that listens on a UDP port (accept) and sends datagrams to that port on an IP address. If the address is broadcast, the behavior is not consistent among platforms tested. This applies whether the broadcast is general (255.255.255.255) or subnet specific (for example, 10.99.255.255)
If the local computer is included in the broadcast specification (which it always is in the general broadcast), the datagram is received on Windows XP and on Mac OS 9.2. But it is not on OS X.
To my thinking the OS X behavior is goofy.
After looking at 'route get' and netstat and ifconfig, I suspect this may be a routing problem (and perhaps involves politics, philosophy, pragmatics and literacy).
If I do a 'route get' on the general broadcast IP address or my local subnet broadcast address, only the ethernet adaptor shows up (en0), the local host does not (lo0).
In the host file 255.255.255.255 is routed to broadcasthost whatever that is. I did notice that a device can have a BROADCAST flag set in ifconfig, so broadcasthost might forward to all devices with that flag set. Device lo0 does not have it set.
The subnet broadcast is handled elsewhere. A netstat -r shows the subnet broadcast routed to en0 and does not include an entry for lo0.
The consistency seems to indicate policy. (A policy that is in violation of the meaning of broadcast addresses.)
So...
I'm still not sure. Can anyone confirm that this is a routing problem?
If so, my strategy is to find a workaround rather than bother RunRev. If I have users fiddle with routing I might break something.
Does this also happen on Linux?
Dar Scott
_______________________________________________ use-revolution mailing list [EMAIL PROTECTED] http://lists.runrev.com/mailman/listinfo/use-revolution
