alex at digriz
Oct 27, 2009, 10:45 AM
Post #2 of 3
(786 views)
Permalink
|
Ted Mittelstaedt <tedm [at] ipinc> wrote:
>
> Tore Anderson wrote:
>>
>> I've been doing some testing in order to determine whether or not it
>> would be «dangerous» for our customers to dualstack their web sites. The
>> largest problem I've found so far affects a very specific group of
>> clients, which:
>>
>> 1) are using Windows Vista or newer, and
>> 2) are using the Opera web browser, and
>> 3) are assigned public IPv4 addresses, and
>> 4) are on a network which filters inbound proto-41 traffic.
>>
>> In this case, the client will have a 6to4 tunnel interface automatically
>> configured, and will prefer using it over native IPv4 for contacting
>> dualstacked web sites.
>
> Further discussion on the SA list about this indicates this is
> a known bug that was reported to Opera - but the broken browsers
> are out there, of course.
>
Eugh, I have been reporting a stupid IPv6 bug for ages (about five years
now) and they seem to ignore me; I was hoping by now they would have
fixed it for version ten. Last time I even got a bug id, unsurprisingly
it has probably been filed under "not important" like all the previous
ones though. :-/
This was not so annoying (it only really ever affected my own sites)
until my IPv6 enabled ISP got themselves whitelisted with Google and now
using it is damn awkward :-/
The bug in detail, under Linux and Windows, it is very noticable that
when you try to access a dual stacked site from an IPv4 only
workstation, Opera will hang without sending a single packet till some
internal timer expires (looks like about five seconds). It's crazy, you
run tcpdump, you see the DNS query and response with an AAAA
record...and then nothing. After a while, Opera decides maybe it should
do something after all.
Christ, even Firefox is not that braindead :-/
</rant>
Time to move on I guess :)
Cheers
--
Alexander Clouter
.sigmonster says: Even bytes get lonely for a little bit.
|
