On 12-okt-2005, at 12:48, Jari Arkko wrote:
What would be the difference between the "second shim" and the actions of the "updated RFC 3484" that we talked about in Amsterdam? Does the latter impact applications? I thought not, but I may be mistaken.
RFC 3484 says that applications should cycle through all the destination addresses it provides. So applications need to do this work. When this is extended to source addresses too the application will have to cycle through all source/dest pairs. This means a significant impact on applications.
If we can put a shim between the sendto, sendmsg, bind and connect API calls and the actual execution of the underlying mechanisms then this can be handled transparently to the application.
A new sendto/sendmsg/bind/connect-by-name API extension could do the same but that of course also require application changes...