Re: [v6ops] Please review the No IPv4 draft

[Simon Perreault <[e-mail]>]

Le 2014-04-15 11:21, Nick Hilliard a écrit :
> On 15/04/2014 14:18, Simon Perreault wrote:
>> Basically we're trying to make DHCPv4 clients shut up.

This was a mistake on my part. We're trying to do much more. We're
trying to create an IPv4 kill switch.

> then there's confusion about the draft's aims.  A good deal of the draft
> goes into detail about completely stopping all ipv4 services on the client
> system (which is something that is not going to be simple to achieve in
> practice), whereas your stated aims and justification in the draft are all
> about getting dhcpv4 clients to stop making noise.
> 
> Getting DHCPv4 clients to shut up is an admirable goal and I think it's
> worth pursuing this by using the front-door method of creating a dhcpv4
> no-service option (suggestion: use dhcpv4, implement a timeout, make it
> interface specific, i.e. don't attempt to apply the configuration to
> multiple interfaces, and don't attempt to complicate things by creating a
> requirement to completely disable ipv4).  This would not be difficult to
> achieve either operationally or from a protocol point of view; it would be
> both clear and unambiguous, and it would not suffer from the technical /
> implementation problems that either Ray or I brought up.  As a side issue
> you may want to consider renaming the draft -nodhcpv4- instead of -noipv4-.
> 
> Getting hosts to stop talking ipv4 completely is a much more difficult
> problem.  If you want to write a draft about this, then you need to start
> from the beginning explaining the problem you're trying to solve,

We do have a section titled "The Problems we're Trying to Solve".

> why you're trying to solve it

Because problems are problematic.

> and then provide guidance on how it should be
> handled.  This analysis is not present in your current draft.

You might be interested in this analysis of problems preventing the
migration to IPv6-only:

http://tools.ietf.org/html/draft-ietf-sunset4-gapanalysis-04

> At the very least I would be curious to know how and why a DHCP or RA
> request coming in on any random interface should have the authority to
> completely disable an entire communications protocol, O/S wide.

It does *not* have any authority. It is signalling. A host is free to
ignore the signalling.

> As a
> client, I might well have multiple interfaces configured, only one of which
> has ipv6 disabled, and it would be difficult to understand how e.g. my 3g
> provider should have the right to tell me that my operating system should
> completely shut down ipv4 while my computer was connected to my home wifi.

It does not. Did you read our draft? We specify how the No-IPv4 option
is to be processed in multiple interfaces settings.

Simon
-- 
DTN made easy, lean, and smart --> http://postellation.viagenie.ca
NAT64/DNS64 open-source        --> http://ecdysis.viagenie.ca
STUN/TURN server               --> http://numb.viagenie.ca