I am not sure what youa re asking with either of your quesitons.
With regard to question 1, you seem to be asking whether we have seen
folks adding EHs, and whether it has been observed to cause
difficulties. We have seen folks trying to standardize exactly such
additions. So I presume they have already implemented them. And we
have seen folks explaining a range of cases where it causes problems.
With regard to 2, you seem to be you seem to be constructing a very odd
reading of an RFC. Some people have clearly said that they read the
existing RFC as permitting additions of EHs. That is not a matter of
future RFCs, but of current readers. Other people have said that they
read the text as clearly prohibiting such behavior. Which would at a
minimum mean that any IETF effort to change it would be required to
explain why it was acceptable and interoperable to change the rules.
Given that the existing wording has been interpreted in different ways
by different people, and that there is good reason to beileve that the
differing interpretations will (if they have not already) cause
interoperability issues, it seems to me incumbent on the WG to be clear
about what it means.
The usual approach to this is to use RFC 2119 terminology and careful
wording.
Yours,
Joel
On 2/17/17 12:15 PM, otroan@xxxxxxxxxxxxx wrote:
Joel,
Given that different people have interpreted 2460 as permitting or prohibiting the addition of Extension Headers by intermediate devices, there clearly is an ambiguity.
I'm a little uncertain if I'm unclear or if you simply didn't read my message. :-)
Do we agree that you can see this at two different angles?
1) Are there any interoperability issue or ambiguity in the protocol specification of 2460 and how implementors of 2460 have interpreted that?
2) Is 2460 "future-ambiguous", i.e it is unclear if 2460 permits a future extension? Like ECMP, Header insertion NAT...
The answer to 1 is no. And no-one has claimed otherwise in these discussions.
For 2, that would be in the area of stating the law for what future extensions of IPv6 can or cannot do.
If we want to go there 2460, yes there are ambiguities. It's hard to predict what the IETF can possible invent in the future and if that should be permitted or not. And of course what effect that would have on future documents regardless.
Clear?
Best regards,
Ole
That is the point that concerns me.
Yours,
Joel
On 2/17/17 9:12 AM, otroan@xxxxxxxxxxxxx wrote:
Fernando,
It is a simple logical consequence.
Middleboxes do not exist in the IPv6 architecture.
There is no interpretation of 2460 that can lead to an implementor inserting headers other places than at the source.
Therefore, there is no interoperability issue in RFC2460 nor any ambiguity that needs to be resolved in RFC2460.
We're not writing law, we're writing interoperable protocol.
Ole
On 17 Feb 2017, at 13:40, Fernando Gont <fgont@xxxxxxxxxxxxxxx> wrote:
On 02/15/2017 07:18 AM, otroan@xxxxxxxxxxxxx wrote:
Ole, it is true that we write in English, and there is always room for
"interpretation", sometimes reasoanble room, sometimes not.
But in this case we have a demonstrated difference in how people
understand the existing text. When we have such a demonstrated
difference, we have an obligation to address it.
This particular issue has caused no interoperability issue,
May I ask what's the data that support this statement?
From the shepherd's writeup:
IPv6 is implemented on most platforms (hosts, routers, servers, etc.),
including proprietary and open source. A list of products that have
received the IPV6 Ready logo can be found at:
https://www.ipv6ready.org/db/index.php/public/?o=4
This has nothing to do wth the interoperability problems that may be
caused by a middlebox that inserts EHs.
You certainly have no way of knowing this, or whether interoperability
issues may arise in the future.
Yes, we do know if our protocols have interoperability issues.
Have you implemented RFC2460? I have. So have many others on this list.
In the context of implementing 2460 there just is no ambiguity and this issue will never arise.
Huh? Yes, if you connect two IPv6 devices, without a middle-box
inserting EHs in the middle, you will not experience the associated
possible problems. What's the news here?
What you are talking about is something else. You are talking about the hypothetical "What if someone standardised something new in the future?"
:-)
C'mon, Ole. Take a look at the initial versions of the SR I-D -- and, EH
insertion has reportedly been deployed as a result of the implementation
of such initial versions of the I-D.
You can clarify that EH insertion is banned, and move rfc2460bis to full
stanard (since that's what's supposed to be mature)
You can delay rfc2460->std, and work to update rfc2460.
Now, moving rfc2460 to full std knowingly leaving a hole there such that
after rfc2460 is std you completely change the architecture (e2e vs
!e2e) with EH insertion doesn't seem a serious thing to do, IMO.
Thanks,
--
Fernando Gont
SI6 Networks
e-mail: fgont@xxxxxxxxxxxxxxx
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492