Yes, “Happy eyeballs” is mainly suggested to something totally different, but
in case NTP client want’s to use only single server as a source, it is fully
applicable for it as well.
If one want’s to use multiple connections, IPv6 will _never_ cause any trouble
if different AF sockets are handled as equals. And if that would be the case
this whole discussion about IPv6 being problematic is absolutely unnecessary.
For the Happy Eyeballs: Note the word _delay_. Delay is added to prevent
unnecessary traffic. If only one connection is needed and it’s successful, the
other AF won’t be used at all. If first AF fails, then we might get reliability
by redundancy. Pretty clear right? Again this is applicable for single
client↔server connections, if some NTP client wanted to do that.
Using v6 for over a decade is actually part of the problem as it gives insights
that tends to be outdated in todays world, as many sensible operators actually
have put effort on it lately. I have also used IPv6 for a long time and have
seen significant improvement in that during the last year or two. And having
two different routes to a destination undoubtedly gives benefit in terms of
availability.
--
Markku Miettinen
From: Mouse
Sent: March 13, 2013 4:49 PM
To: [email protected]
Subject: Re: [Pool] Getting close to 1000 IPv6 servers
> Second of all, the IETF has worked on e.g. something called â??Happy Eyeball$
Your sketch of it makes it sound as though this is designed fairly
specifically for Web use; even the name makes it sound as though it's
all about presenting data to a human.
As described, it is completely inapplicable to something that, like
NTP, does not establish connections. NTP could do something similar,
using multiple addresses (if returned) for a peer specified by name;
personally, I'd call that unnecessary complication.
> Client launches both v6 and v4 sockets with reasonable delay to prevent addi$
You're saying you start multiple connections every time to..._reduce_
traffic? Run that by me again?
> In my experience IPv6 outages are way over exaggerated, there are similar th$
That disagrees with my experience. I've been using both protocols for
over a decade, and my experience is that v6 outages are substantially
commoner than v4 outages - and that v4 outages are almost invariably
accompanied by v6 outages as well. I can't offhand think of even a
single incident when v4 was broken but v6 was working (except by
design, when a piece of network deliberately has a v6-only uplink).
/~\ The ASCII Mouse
\ / Ribbon Campaign
X Against HTML [email protected]
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool
