[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: v6ops-v6onbydefault: link-locals and AI_ADDRCONFIG

On Wednesday 29 October 2003 20:05, Pekka Savola wrote:
> On Wed, 29 Oct 2003, Juan Rodriguez Hervella wrote:
> > On the other hand, I think that applications which
> > query the DNS shouldn't be real-time applications, so I see this
> > feature is not very useful.
> How do you see people browsing the web get to the websites then, if not
> using DNS?  That's enough realtime in my book -- the user waits while the
> request(s) is being processed.
> I really don't understand what you're trying to say..

Hello Pekka!

I will try to explain it better....

I just say that this option is not *very* useful, having
into account the problems that it has with link-local
addresses and so on... if you fix these problems, it's
a matter of choice, if you want to use it, just do it.

> > If I understood Stig well, he said that people was
> > turning off IPv6 because of delaying issues. Doing that is not an option
> > if you want to allow your app. to run over "IPv6 AND IPv4". If you can
> > not cope with such delays, maybe you shouldn't be using the DNS at all.
> Then what?  Distributed hosts file? ;-) ?

 ^--^ ...(after a while)... pouting (grrrr).

Then what ? it's not my problem, I haven't switched to IPv4 because
the IPv6 resolver took a couple of extra roundtrip times.

I'm happy with a lot of IPv6 applications that don't
use this feature at all.

> > The problem here is that most of the times this feature will fail back
> > to the worst case, which is the normal getaddrinfo behaviour, because
> > it's not as easy as checking for global IPv6 configured address, as
> > Itojun has already said. We might come up with a lot of scenarios where
> > this option might fail to improve the response time (e.g: only link-local
> > addresses, global addresses but no default route, proper link local IPv6
> > configuration but some black-hole on outer routers.....).
> I fail to see the point here:
>  - if only link local addresses: fixed AI_ADDRCONFIG would fix
>  - if global but no default route:
>     a) without the on-link assumption: automatic immediate no route
>        message and fallback to the next address
>     b) with (current) on-link assumption: problematic (which is why the
>        assumption must be killed :-)

Uhm I thought that IPv6 was a piece of cake, but it might be problematic
sometimes....Ops, I've got the solution, just kill IPv6 ! ;-)

>  - if global, default route, but blackhole somewhere:
>       silently discarding packets is evil practice anyway and nothing can
>       be done about that, whether in case of v4 or v6.
>  - if global, default route, everything works, but some DNS server munges
>       the AAAA queries: can't help with that, but because v6 would be
>       used, that has to be fixed by fixing the servers anyway.
> .. so my perception is that this would fix all the problems related to
> partial IPv6 deployment I could think of w/ the elimination of the on-link
> assumption pretty well.

I see another way of fixing:

If this option is problematic, don't use it. This has been proved to work,
as a lot of applications don't use it already.

You will have to give me other arguments to kill the on-link assuption.


IETF IPv6 working group mailing list
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6