Re: [Softwires] No change to RFC6145 needed for 4rd-U

[Rémi Després <[e-mail]>]

Le 2012-02-02 à 11:59, Maoke a écrit :

> dear Remi,
> 
> sorry but your change of the title of this thread is misleading. i mean: no change to RFC6145 needed for practice of operation. my observation on the ICMP transparency of RFC6145 in double translation is relatively independent, without much relationship to 4rd-U.


> well, the only connection is: 4rd-U REQUIRES IPv6 carrying ICMPv4 messages without translation. 

4rd-U is indeed proposed with all IPv4 payloads transparently transmitted, including ICMPv4.
This could be changed if ICMPv4 if there would be a convincing need.
Honestly, what you have provided so far hasn't been convincing. 


> 2012/2/2 Rémi Després <[e-mail]>
> 
> Le 2012-02-01 à 03:04, Maoke a écrit :
>> ...
>> 
>> i have investigated, technically, what if we had updated RFC6145 with carrying ICMPv4 messages in IPv6 directly instead of translating to ICMPv6, specially for double translation,
> 
> Introducing ICMPv4 messages in IPv6 packets "instead of" ICMPv6 messages in RFC6145 would be an incompatible change. 
> I don't think anyone proposed that.
>  
> What might be envisaged without the same backward incompatibility, is that IPv6-only hosts would accept in IPv6 packets BOTH IPv6 ICMP messages and ICMPv4 messages.
> 
> Yet, it remains to be analyzed whether it would be useful in real world.
> 
> This depends on a detailed use case being identified, where double RFC6145 translation would have a substantial advantage over the header-mapping variant of 4rd-U-03 (4rd-H). (This advantage would have to compensate for the loss of IPv4 transparency as good as that of 4rd-E, and for the loss of the tunnel-specific traffic class which can be used in 4rd-E.) 
> 
> This use case, key for this discussion, is subject of another e-mail thread.
> 
> i (and others) have tested the single/double-compatibility in practice and confirmed it is useful at least for some users.

I don't deny. I just believe that, to have a chance to be convinced, I need a description of at least a use case where it is useful (with applicable mapping rules and indication of whether there is a DNS64).

> another mail thread might fall into pure philosophical debate, which is not preferred,

Whatever the thread, I look for a technical discussion, not a philosophical one.

RD



 
>  
> 
> - maoke 
>  
> 
> RD
>