On Apr 13, 2011, at 11:23 PM, Dmitry Anipko wrote:

> Hello Mark,
> 
> I agree that if the 4rd IPv4 address is only used in such a way, that 
> applications cannot start use the address directly (such as e.g. perform 
> binding to the address assigned to an interface), then this issue doesn't 
> apply.
> 
> However I'm not sure that the current draft makes clear that this is the only 
> intended use. For example, it defines CE as "a node.. it may be a host..", 
> instead of saying that CE is a function implemented in a host. Also, the 
> introduction in section 1 also does not include NAPT as a required element of 
> the solution (it says "The 4rd mechanism tunnels IPv4 over IPv6 using an 
> algorithmic mapping.."). 
> 
> Some existing implementations may be assigning the address used for NAPT to 
> an interface (and as long as the address is for the host exclusive use, an 
> implementation is able ensure such usage doesn't cause problems) - that's 
> what Windows ICS you mentioned actually does.
> 
> With the current language, I don't think it is unlikely that implementers 
> will be tempted to assign a 4rd IPv4 address to an interface (e.g. with 
> intent to enable wider set of apps to work without a NAPT in the middle), and 
> if the spirit of the draft is that the IPv4 address assigned to the node by 
> 4rd should be only be used for CE function, performing NAPT, (or more 
> specifically, should not be made available for applications to bind) then I 
> think it may be useful to state it as such to guide implementers.

Then at least you and I agree. If 4rd becomes a WG item, this is the kind of 
change we should be able to affect in the draft with WG consensus. 

- Mark

> 
> Thank you,
> Dmitry
> 
> -----Original Message-----
> From: Mark Townsley [mailto:[email protected]] 
> Sent: Tuesday, April 12, 2011 1:04 PM
> To: Dmitry Anipko
> Cc: [email protected]
> Subject: Re: [Softwires] sharing restricted addresses by hosts in 4rd 
> (draft-despres-intarea-4rd-01)
> 
> 
> Hello Dmitry,
> 
> My view is that 4rd is most easily understood if and only if it connects to a 
> CE function that is performing NAPT. The CE function may be in what is 
> traditionally considered a host, or in what is clearly a router.
> 
> More specifically, a device that is forwarding packets from one interface 
> (virtual or otherwise) to another through a NAPT that has one interface with 
> IPv6 configured (via DHCPv6 or otherwise) as performing 4rd (which enables 
> dual-stack via a port-restricted IPv4 address for the NAPT using IPv6 as the 
> transport) then you a have a 4rd CE. That could be a "host" in that it is a 
> Windows PC with internet connection sharing for IPv4 turned on and hence 
> forwards packets between interfaces with a NAPT due to the IPv4-enabled 
> interface created when 4rd is configured. 
> 
> I would avoid anything that requires the host forwarding table to be altered 
> to accommodate 4rd. Instead, the NAPT function that is already present in a 
> small router or host configured to look like a router is modified to use a 
> set of ports that it is allowed to use when 4rd is enabled. 
> 
> - Mark
> 

_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to