Re: problem statement [was Re: New Version Notification for draft-hinden-ipv4flag-00.txt]

[Nick Hilliard <[e-mail]>]

Mikael Abrahamsson wrote:
> Indeed. And I'd rather put this hack in IPv6, as it requires adding code
> to configure and send an additional RA option compared to requiring a
> complete IPv4 stack and DHCPv4 code to send some option there.

I agree it will simplify things on the operator side, slightly.  The
complication on the client side is substantial though, and it's usually
the client side where we run into deployment problems due to high levels
of heterogeneity and in this case, layering issues, i.e. using ipv6 to
signal to a kernel which needs to message-pass to a dhcpv4 userland
process to stop issuing queries on a specific interface or to otherwise
implement what we mean by this flag.  There would also be a requirement
for the network edge to be able to filter out RAs with this option,
given that the ipv6 / RA model is to be able to deploy multiple routers
and we cannot assume that this flag would be limited to e.g. cellular
networks or resi access networks or whatever.  This added complication
is why I question the wisdom of the RA approach.  Simple, even if it
doesn't solve everything, is usually better.

Nick