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

[Nick Hilliard <[e-mail]>]

Mikael Abrahamsson wrote:
> This functionality doesn't exist today in any access network I know of.
> It's also non-trivial to do considering how access networks are done.

Both of these statements also apply to encoding a no-ipv4 flag in an RA
message.

> Preferrably one would like to ahve an ethertype filter on the access
> port to only allow 0x86dd and nothing else. Then putting this flag into
> IPv4 means this isn't possible.

if you can filter everything except 0x86dd (or the equivalent for
different media, e.g. docsis, pppoe, etc), then the ipv4 problem
statement which draft-hinden-ipv4flag is intended to solve has already
been handled.

> IPv6 is the next generation IP protocol. I prefer that it can signal
> that the legacy version is not available, instead of having to keep
> legacy IP functionality around.
> 
> ... or we put it in some other L2 protocol, such as LLDP.

It comes down where in your stack you want to put your hacks.  A sentry
dhcpv4 agent is simple to define in terms of protocol scope, behaviour
and meaning, and has relatively well constrained requirements in terms
of the amount of ipv4 infrastructure required to sustain it (virtually
none, in the minimal case).

If you want to define and implement this functionality via some other
protocol, it requires potentially substantial complication on the client
side, as well as creating a requirement to define what is meant by "no
ipv4".  This, IIRC, was one of the issues that led to the sunset4 draft
not progressing further.

Nick