Re: [homenet] Please review the No IPv4 draft

2014-04-18 Thread Philip Homburg
In your letter dated Thu, 17 Apr 2014 22:16:19 -0430 you wrote:
  I like the idea of having the No-IPv4 option for DHCPv6. I don't like
the idea of having this option in a v4 world; mainly because in mix
networks where you might have some folks which only speak IPv4, imagine
a case where this option is sent by a DHCPv4 to a client who do not
understand v6. Of course, probably the client won't understand the
option and won't turn off IPv4. There is also the chance that this is a
modern OS but for some reason IPv6 was disable in the host, yes, it will
understand the option No-VP6 (sent by a dhcp4).

As far as I understand, this draft addresses the situation where the network
doesn't offer IPv4 in the first place.

So any IPv4-only host would have no internet connectivity anyhow, whether it
processes the option or not.


___
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet


Re: [homenet] Please review the No IPv4 draft

2014-04-18 Thread Simon Perreault
Le 2014-04-17 17:14, Mark Andrews a écrit :
 0 full IPv4 connectivity
 1 all IPv4 off on the interface
 2 local (link/site not global) connectivity only on the interface
 3 no IPv4 on the machine.

 Uh, no. The levels are in strictly increasing order. 2 includes
 everything in 1, plus some more:

2 - No IPv4 upstream, local IPv4 restricted:  Same semantics as value
   1, with the following additions:

 2 deals with propagation to downstream interfaces. It has nothing to do
 with link or site local stuff.
 
 Part of the problem here is that upstream is the only interface
 for singlely attached devices other than loopback.  There is no
 only local connectivity is available step for such nodes.

Correct. For such nodes, 1 == 2. Is that a problem? It's still a
stairway going up, with steps 1 and 2 being at the same height for a
subset of nodes.

Simon

 
0 - IPv4 fully enabled:  This is equivalent to the absence of the No-
   IPv4 option.  It is included here so that a DHCPv6 server can
   explicitly re-enable IPv4 access by including it in a Reply
   message following a Reconfigure, or similarly by a router in a
   spontaneous Router Advertisement.
 
1 - No IPv4 upstream:  Any kind of IPv4 connectivity is unavailable
   on the link on which the option is received.  Therefore, any
   attempts to provision IPv4 by the host or to use IPv4 in any
   fashion, on that link, will be useless.  IPv4 MAY be dropped,
   blocked, or otherwise ignored on that link.
 
 Mark


-- 
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

___
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet


Re: [homenet] Please review the No IPv4 draft

2014-04-18 Thread Simon Perreault
Le 2014-04-17 23:49, Lorenzo Colitti a écrit :
 On Thu, Apr 17, 2014 at 11:24 PM, Simon Perreault
 simon.perrea...@viagenie.ca mailto:simon.perrea...@viagenie.ca wrote:
 
 Le 2014-04-16 12:25, Simon Kelley a écrit :
  Android uses dhcpcd as its DHCP client.
 
 Thought about this some more... Doesn't dhcpcd implement everything in a
 single daemon? i.e., DHCPv4, DHCPv4, IPv4LL, IPv6 RS? If so, then it
 would be easy for one protocol to affect another, wouldn't it?
 
 
 In practice, you can't use dhcpcd for autoconf because it doesn't
 support privacy addresses or removing the default route when the RA
 lifetime expires. (Or at least, the 6.3.2 version I downloaded today and
 ran on Debian stable didn't seem to support this).
 
 So you still need to run kernel autoconf, at which point you might as
 well disable dhdpcd's RS processing since you probably don't want both
 the kernel and dhcpcd to be running autoconf.

Got it. So, summarizing, for Android, DHCPv4 and DHCPv6 options would
likely not be problematic, but an RA option likely would.

Thanks,
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

___
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet


Re: [homenet] Please review the No IPv4 draft

2014-04-18 Thread Alejandro Acosta
Hi Philip,

El 4/18/2014 3:28 AM, Philip Homburg escribió:
 In your letter dated Thu, 17 Apr 2014 22:16:19 -0430 you wrote:
  I like the idea of having the No-IPv4 option for DHCPv6. I don't like
 the idea of having this option in a v4 world; mainly because in mix
 networks where you might have some folks which only speak IPv4, imagine
 a case where this option is sent by a DHCPv4 to a client who do not
 understand v6. Of course, probably the client won't understand the
 option and won't turn off IPv4. There is also the chance that this is a
 modern OS but for some reason IPv6 was disable in the host, yes, it will
 understand the option No-VP6 (sent by a dhcp4).
 
 As far as I understand, this draft addresses the situation where the network
 doesn't offer IPv4 in the first place.

You are absolutely right, thanks.
But anyway the scenario I mentioned above can also exists (and for good
or bad eventually will happen).. I just wanted to express my position
about the option NO-IPv4 in DHCPv4

 
 So any IPv4-only host would have no internet connectivity anyhow, whether it
 processes the option or not.
 
 

___
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet


Re: [homenet] Please review the No IPv4 draft

2014-04-18 Thread Ted Lemon
On Apr 18, 2014, at 7:51 AM, Simon Perreault simon.perrea...@viagenie.ca 
wrote:
 Got it. So, summarizing, for Android, DHCPv4 and DHCPv6 options would
 likely not be problematic, but an RA option likely would.

This is weird, though.   How does a DHCPv6 client know to attempt 
configuration, if it doesn't see RAs?   I just hacked around this on my router 
at home in order to get Comcast IPv6 working, but that seems broken.

___
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet


Re: [homenet] Please review the No IPv4 draft

2014-04-18 Thread Simon Perreault
Le 2014-04-18 11:56, Ted Lemon a écrit :
 On Apr 18, 2014, at 7:51 AM, Simon Perreault simon.perrea...@viagenie.ca 
 wrote:
 Got it. So, summarizing, for Android, DHCPv4 and DHCPv6 options would
 likely not be problematic, but an RA option likely would.
 
 This is weird, though.   How does a DHCPv6 client know to attempt 
 configuration, if it doesn't see RAs?   I just hacked around this on my 
 router at home in order to get Comcast IPv6 working, but that seems broken.

I don't know how Android does it, but any DHCPv6 client is free to
unilaterally attempt configuration. It doesn't need to wait for an RA to
tell it to do so. RAs are just a hint.

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

___
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet