RE: network controls are necessary

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Fri, 6 Dec 2002, Joel M. Halpern wrote:

> If I thought having the network make the selections would require change in
> application logic, I would not be suggesting such a solution.
> I would prefer a solution that is completely invisible to the application
> (such as Tony Li's suggesting that the application is handed a 32 bit
> identifier when it currently expects an IP address).
> Failing that, I would prefer to minimize the application coupling.
> Having application level address selection increases the application impact.

A few weeks back, on the IETF discussion list in the "kernelizing the
address lookup" (or similar) thread, the opposite point was made, by
several people if I remember correctly.

Some applications need to see more than just the host name. If we are
going to build something where a host can have several addresses tied to
several paths with different properties (fast/slow, secure/insecure,
free/cheap/expensive), applications will want to make selecting
addresses their business.

Currently, IPv6 doesn't support backwards compatibility with
applications that use the traditional socket API. The past three days
I've been trying to find a web browser with IPv6 support for my new
iBook, so I think I've earned the right to say there aren't enough IPv6
applications yet that breaking those is a big issue. However, if we go
down the path of requiring applications to change, we should make very,
very, very sure this is a one time thing and we build in everything we
need for IPv6 - IPv15.