Thus spake <john(_dot_)loughney(_at_)nokia(_dot_)com>
Already we are seeing laptops with multiple interfaces (Ethernet,
WLAN, Bluetooth, etc.) - I think it is only a matter of time that
applications start to take advantage of this.
I see it as a failure of the network layer if applications need to
distinguish between multiple interfaces or address scopes.
While many don't see the distinction, this is the kind of stuff that
desperately needs to be in the OS/stack -- if anywhere. Applications should
not know about network-level details unless they have a specific reason to
care, and if anything more than a small majority of applications care, there
is something wrong with our model.
Most applications today cannot even handle the case of multiple A records
for a destination host; do we really expect _tens of thousands_ of
applications to be updated with source-address-picking intelligence?
A lot of this is OS / implementation specific stuff, but I don't
understand
why this should not be supported in standards.
The approach to other problems, such as QOS, is to provide a common
architecture and terminology, then let the individual vendors provide their
own (often creative) solutions within those confines.
S
Stephen Sprunk "God does not play dice." --Albert Einstein
CCIE #3723 "God is an inveterate gambler, and He throws the
K5SSS dice at every possible opportunity." --Stephen Hawking