Re: [v6ops] The need for local-ipv4 socket transition solutions -- NAT64/DNS64 remains insufficient

[James Woodyatt <[e-mail]>]

Lorenzo—

Let me put it this way. Until someone [you?] writes this draft and gets it
into the pipeline at 6MAN, this group won't have any real idea of precisely
what is the CLAT that it might consider recommending be implemented in All
The IPv6 Hosts. Maybe you don't want to go through POSIX.1 looking for all
the corner cases when you're writing the draft, but I would recommend you
do that anyway. Someone will, and you can either have the answers ready at
the start, or you can have an endless discussion about the unanswered
questions. I'm sure I can think of lots and lots and lots of questions that
should be easy for someone who has implemented a CLAT to answer.


On Thu, Apr 23, 2015 at 9:00 PM, Lorenzo Colitti <[e-mail]> wrote:

> On Thu, Apr 23, 2015 at 7:36 AM, James Woodyatt <[e-mail]> wrote:
>
>> In particular, I think it's critical to have a clear understanding of
>> precisely what features of the PF_INET interfaces are expected to be
>> handled by the CLAT function. If the IP_PKTINFO socket option, for example,
>> is expected to be unsupported, then it's important to know that. What
>> does the CLAT do with IPv4 link-local scope addresses? What about the
>> SO_REUSEADDR option? A good place to start would probably be to go through
>> the entire POSIX.1 standard, and for all the IPv4 networking interfaces,
>> describe what is and isn't supported by the CLAT function recommended by
>> IETF.
>>
>
> I don't think such a fine-toothed-comb endeavour makes sense. 464xlat is
> just an unnumbered point-to-point link. That directly implies that the
> answers to your questions are "there's no reason not to support IP_PKTINFO,
> and no extra work required to do so", "link-local scope addresses are not
> supported because it's an unnumbered point-to-point link", and
> "SO_REUSEADDR means exactly what it means on any other IP address on the
> system".
>
> I suppose we can always write a document saying the above, plus
> recommending that there be a different local clat IP address for every
> interface on the system. Such a document wouldn't be very long, but I
> suppose it might be helpful.
>
> I don't think there is a need to go through POSIX.1. looking for corner
> cases.
>



-- 
james woodyatt <[e-mail]>
Nest Labs, Communications Engineering