On Thu, 27 Oct 2011, The IESG wrote:
Hi,
The IESG has received a request from the IPv6 Operations WG (v6ops) to
consider the following document:
- 'Happy Eyeballs: Success with Dual-Stack Hosts'
<draft-ietf-v6ops-happy-eyeballs-05.txt> as a Proposed Standard
Having been on a lot of different (company and few end user) networks the
last weeks that had IPv6 and having seen the discussion on v6ops only just
today I want to second a couple of voices from v6ops (sorry for the
lateness and having to do it this way) and raise mine as well on
:: 5.3. Debugging and Troubleshooting
:: ..
:: To assist in that regard, the
:: implementations MAY also provide a mechanism to disable their Happy
:: Eyeballs behavior via a user setting.
here. ++ on MAY is NOT strong enough, especially given the ignorance of
vendors for adding such an option voluntarily anyway. I understand that
there are situations in which a MUST is too hard and not feasible but
here's a few observations:
- Given I try to be v6-only whenever possible I see v6 problems (HE) users
on v4/v6 networks do not see usually fairly quickly.
- As soon as having machines on dual-stack, observations are:
a) a significant drop in use of v6-favoured transition technology (e.g.
NAT64 traffic) when comparing HE vs. non-HE systems by "more
'happy eyeballs' enforced" use of legacy IP.
b) I find that about about 1 in 3 of IPv6 enabled networks have problems
that are simply hidden to most people by HE these days, and may
otherwise not be enough to bother and ignored, and it is extremly hard
to
i) show people the problem exists (send them to a v6-only server is
the fastest solution usally unless (v) applies)
ii) then explain it to them (as things are mostly so fine lately)
iii) for them then to figure out when the problem started
iv) for them to figure out where the problem is and what is
affected, and
v) sometimes it is strange problems only affecting some but not
all systems making debugging really hard.
c) that debugging often takes long (often due to missing options, etc),
d) that monitoring is usualy far from existent to catch the problems
early or at all, and
e) by the time it is understood and fixed some other "minor
oddities here and there" strangely often disappear as well and
things are indeed "so fine" and use some more v6 in addition as well.
I have about too many emails on these topics in my inbox from the last
weeks (and not just with "you") and I am tired of not being able to turn
HE off on a good fraction mainstream systems/browsers, especially for the
people who SHOULD, want, or rather MUST care (contrary to the people who
have no idea what this would be about and never bother the hard to find
advanced option anyway).
I vote for substituing "MAY" to a "SHOULD whenever possible" in 5.3.
/bz
--
Bjoern A. Zeeb Damn eat your dogfood! Really!
Stop bit received. Insert coin for new address family.
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf