Re: [exim-dev] IPv6 reachability of tahini

On 2010-09-06 at 14:16 +0200, Jakob Hirsch wrote:
> Phil Pennock, 2010-09-06 00:34:
>
> > I've now found that I can ssh to tahini just fine.
>
> Also bulk transfers (scp etc.)? If not, it may be a MTU issue (SixXs
> recommends 1280, AFAIR).

ssh worked, git over ssh did not, so I'm suspecting MTU too, even though
I use the recommended 1280.

I've only noticed this problem with tahini. I use IPv6 *extensively* on
my colo box and have no problems reaching it. That's native IPv6.

I also did not have problems like this with an HE.net tunnel. Instead,
I had instability which killed long-lived TCP connections (SSH to my
colo-box) which is why I switched to SixXS.

Short version seems to be: tunnelling sucks.

> Does the "telnet test" Graeme used works for you?

Shockingly enough, it's one of the first things I did to try to track
down a problem; alas, I failed to remember it when writing my mail.
Since TCP is established, it's clearly not a regular ACL issue; perhaps
something filtering ICMPv6 ...

> > So it's only port 80 that's having troubles over IPv6 from a SixXS
> > end-point.
>
> I could probably try that tonight.

No, looks like it's on my end. I need to find a non-broken IPv6 setup
for home. Looks like I'm going to have to go back to 6to4, which is
hard to debug and unpredictable, but which ironically has given me far
fewer problems than "proper" tunnels. And hey, my ISP has finally clued
in and started running 6to4 gateways themselves and they have native
IPv6 peering, I believe.

Sorry for the noise.
-Phil