Strange socket bind behavior on multihome system

Hi guys,

On a multihome Windows system I am sending packets over a raw socket. If I bind to an IP address with a higher metric in the route table, the packet actually goes out an interface with a lower metric in the route table. The IP src addr in the packet is still the bound IP address. I think I am hitting the route feature as described in the following.

blogs.msdn.com/zhengpei/archive/2007/04/25/socket-binding-on-a-multihomed-host.aspx

Quoted from above link:

"Even if you call bind() to bind the socket to a specific local address (which means connection manager's AutoLSP is not associated with the socket), still your packets on this socket are subject to the routing decision in terms of choosing which physical interface will be used. "

If I change the route to make the bound IP address a lower (therefore preferred) metric then it works as expected.

Does anyone know if this is the case? Is there anyway around this? I'm trying to use Pcap to grab packets and am finding it impossible since my packets are going out a different interface than I expect!

Thanks,

Mona

this would be a networking question methinks, unless you have code ...