Re: COntrolled-domains (was Re: Alvaro Retana's Discuss on draft-ietf-6man-rfc2460bis-09: (with DISCUSS and COMMENT))

["Stefano Previdi (sprevidi)" <[e-mail]>]

> On Apr 11, 2017, at 5:33 PM, Fernando Gont <[e-mail]> wrote:
> 
> On 04/11/2017 02:39 PM, Brian E Carpenter wrote:
>> On 11/04/2017 19:11, Stefano Previdi (sprevidi) wrote:
>>> 
>>>> On Apr 10, 2017, at 10:24 PM, Brian E Carpenter <[e-mail]> wrote:
>>>> 
>>>> Excuse front posting, but the interleaving is getting a bit distracting in itself.
>>>> 
>>>> My considered opinion is that pretending to forbid "examine" is just laughable,
>>>> since we know that middleboxes make it their job to examine packet headers. 
>>>> So it's pointless to state it; take it out (but leave in the reference to 7045).
>>>> 
>>>> The word "process" is too vague - and has been too vague since RFC 1883.
>>>> So take that out too. (Clearly, a middlebox that can examine a header
>>>> can process it - but so what? That doesn't affect the contents of the
>>>> packet.)
>>>> 
>>>> What we can meaningfully prohibit in the context of Internet-wide interop
>>>> is "insert, modify or delete”.
>>> 
>>> 
>>> we don’t have a definition of “internet-wide” and we may well end up in another loop tying to agree what it means.
>>> 
>>> The simplest approach is to add an exception which consists of a controlled/closed domain (for which we do have a definition).
>> 
>> Unfortunately, we don't, as I discovered when I proposed such an
>> approach to allow local use of the IPv6 Flow Label. "Internet-wide"
>> on the other hand seems completely clear to me. Any node to any node.
> 
> My understanding is that the "controlled domains" that R people talk
> about is a different thing: they talk about controlled domains regarding
> who inserts/removes EHs, rather than src/dst addresses of th packets.
> 
> If packets still have internet-wide srd/dst addresses, then packets
> require internet-wide interoperability.


as described in draft-voyer-6man-extension-header-insertion, the controlled domain is the domain where the packet is sourced _and_ (please, note the “and”) destined.

IOW, both SA and DA belong to the domain under the same control and within that domain all boxes that the packet will traverse are also under the same control.

draft-voyer-6man-extension-header-insertion also gives a couple of examples.

s.

> 
> P.S.: This is a side comment -- as noted before, I'm against
> "applicability statements" on rfc2460 for a bunch of reasons.
> 
> Thanks,
> -- 
> Fernando Gont
> SI6 Networks
> e-mail: [e-mail]
> PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492
> 
> 
> 
>