IPv6 now completely broken in Opera 10.10
Christof Meerwald
completely broken - previously it was only half broken (it didn't fallback
from IPv6 to IPv4 if there was any problem), but now it doesn't even try
IPv6 any more - not even if the site is IPv6 only.
opera -debugdns shows:
dns: Resolving 'www.sixxs.net'
dns: Trying IPv4 lookup for host 'www.sixxs.net'...
dns: Thread created
dns: gethostbyname succeeded
dns: Host name lookup completed with error code 0
dns: Host 'www.sixxs.net' resolved to 213.204.193.2
dns: Resolving 'ipv6.idnet.net'
dns: Trying IPv4 lookup for host 'ipv6.idnet.net'...
dns: Thread created
dns: gethostbyname failed with return value 0 (Success). hostent pointer:(nil)
BTW, I do have native IPv6 here and other browsers are happy to work over
IPv6.
This really seems strange - the world is slowly moving towards IPv6, but
Opera is making a step back to IPv4 only...
Christof
--
http://cmeerw.org sip:cmeerw at cmeerw.org
mailto:cmeerw at cmeerw.org xmpp:cmeerw at cmeerw.org
Bob Goddard
> Just noticed that with Opera 10.10 (on Ubuntu 9.10 AMD64), IPv6 is now
> completely broken - previously it was only half broken (it didn't
> fallback from IPv6 to IPv4 if there was any problem), but now it
> doesn't even try IPv6 any more - not even if the site is IPv6 only.
>
> opera -debugdns shows:
>
> dns: Resolving 'www.sixxs.net'
> dns: Trying IPv4 lookup for host 'www.sixxs.net'...
> dns: Thread created
> dns: gethostbyname succeeded
> dns: Host name lookup completed with error code 0
> dns: Host 'www.sixxs.net' resolved to 213.204.193.2
> dns: Resolving 'ipv6.idnet.net'
> dns: Trying IPv4 lookup for host 'ipv6.idnet.net'...
> dns: Thread created
> dns: gethostbyname failed with return value 0 (Success). hostent
> pointer:(nil)
>
>
> BTW, I do have native IPv6 here and other browsers are happy to work
> over IPv6.
>
> This really seems strange - the world is slowly moving towards IPv6,
> but Opera is making a step back to IPv4 only...
Good grief, you're right! I'd like it back as well as it's about the
only way providers can gauge usage.
B
Eirik Byrkjeflot Anonsen
> Just noticed that with Opera 10.10 (on Ubuntu 9.10 AMD64), IPv6 is now
> completely broken - previously it was only half broken (it didn't fallback
> from IPv6 to IPv4 if there was any problem), but now it doesn't even try
> IPv6 any more - not even if the site is IPv6 only.
>
> opera -debugdns shows:
>
> dns: Resolving 'www.sixxs.net'
> dns: Trying IPv4 lookup for host 'www.sixxs.net'...
> dns: Thread created
> dns: gethostbyname succeeded
> dns: Host name lookup completed with error code 0
> dns: Host 'www.sixxs.net' resolved to 213.204.193.2
> dns: Resolving 'ipv6.idnet.net'
> dns: Trying IPv4 lookup for host 'ipv6.idnet.net'...
> dns: Thread created
> dns: gethostbyname failed with return value 0 (Success). hostent pointer:(nil)
>
>
> BTW, I do have native IPv6 here and other browsers are happy to work over
> IPv6.
>
> This really seems strange - the world is slowly moving towards IPv6, but
> Opera is making a step back to IPv4 only...
Sounds like a bug, please file a bug report
(https://bugs.opera.com/wizard/).
Of course, over the years since we implemented ipv6 support in opera for
linux (back in 2002 or 2003, I think) we have generally moved towards
trying to avoid using it. Ipv4 works pretty much everywhere, ipv6 still
works almost nowhere. And the way things fail tend to give us problems
like failing to fall back to ipv4 when ipv6 failed. (Some of that is
due to architectural issues within opera. Some of it is due to the way
ipv6 fails. Details are unclear to me, as I haven't been working on
this.)
eirik
Christof Meerwald
> Sounds like a bug, please file a bug report
> (https://bugs.opera.com/wizard/).
Just done that now: DSK-272658, let's see if anything happens...
> Of course, over the years since we implemented ipv6 support in opera for
> linux (back in 2002 or 2003, I think) we have generally moved towards
> trying to avoid using it. Ipv4 works pretty much everywhere, ipv6 still
> works almost nowhere. And the way things fail tend to give us problems
> like failing to fall back to ipv4 when ipv6 failed. (Some of that is
> due to architectural issues within opera. Some of it is due to the way
> ipv6 fails. Details are unclear to me, as I haven't been working on
> this.)
While I understand that it can sometimes be difficult to figure out when
IPv6 can (or should be use), pretty much all other browsers seem to have
found a working solution.
Creideiki
> On Tue, 01 Dec 2009 09:34:44 +0100, Eirik Byrkjeflot Anonsen wrote:
>> Sounds like a bug, please file a bug report
>> (https://bugs.opera.com/wizard/).
>
> Just done that now: DSK-272658, let's see if anything happens...
According to <URL:http://bugs.gentoo.org/289111>, it's been in Opera's
bug tracker as DSK-268092 since at least 10.10_pre4672. I first
noticed it in 10.10_pre4720, and am still highly annoyed by it in
10.10 final.
richard lucassen
> Of course, over the years since we implemented ipv6 support in opera
> for linux (back in 2002 or 2003, I think) we have generally moved
> towards trying to avoid using it. Ipv4 works pretty much everywhere,
> ipv6 still works almost nowhere.
Welcome to the real world:
http://www.inetcore.com/project/ipv4ec/index_en.html
It's not a question of "ipv4 works pretty much everywhere, ipv6 still
works almost nowhere". Maybe that was true a few yaers ago, but nowadays
the ipv6 network is getting more and more mature. Avoiding ipv6 in 2009
is ostrich policy.
R.
--
___________________________________________________________________
It is better to remain silent and be thought a fool, than to speak
aloud and remove all doubt.
+------------------------------------------------------------------+
| Richard Lucassen, Utrecht |
| Public key and email address: |
| http://www.lucassen.org/mail-pubkey.html |
+------------------------------------------------------------------+
Eirik Byrkjeflot Anonsen
> On Tue, 01 Dec 2009 09:34:44 +0100
> Eirik Byrkjeflot Anonsen <ei...@opera.com> wrote:
>
>> Of course, over the years since we implemented ipv6 support in opera
>> for linux (back in 2002 or 2003, I think) we have generally moved
>> towards trying to avoid using it. Ipv4 works pretty much everywhere,
>> ipv6 still works almost nowhere.
>
> Welcome to the real world:
>
> http://www.inetcore.com/project/ipv4ec/index_en.html
>
> It's not a question of "ipv4 works pretty much everywhere, ipv6 still
> works almost nowhere". Maybe that was true a few yaers ago, but nowadays
> the ipv6 network is getting more and more mature. Avoiding ipv6 in 2009
> is ostrich policy.
Maybe. But we're not avoiding ipv6 just to avoid it. We've just been
turning off ipv6 features in response to bug reports that shows us cases
where using ipv6 would make opera fail completely. As long as ipv4
actually works, it is better to successfully use ipv4 than to fail to
use ipv6. Of course, if ipv4 no longer works everywhere, then we will
have to live with failing completely in some cases. We just have to
pick our poison.
Disclaimer: I don't actually work on this stuff, so I know very little
about how our ipv6 vs ipv4 situation really is today. And most likely
the information I do have is somewhat out of date.
eirik
Eirik Byrkjeflot Anonsen
Ok, according to that bug report we have recently fixed at least one
cause for this. Hopefully the problem will be gone in the next update.
eirik
Bob Goddard
> richard lucassen <spam...@xaq.nl> writes:
>
>> On Tue, 01 Dec 2009 09:34:44 +0100
>> Eirik Byrkjeflot Anonsen <ei...@opera.com> wrote:
>>
>>> Of course, over the years since we implemented ipv6 support in opera
>>> for linux (back in 2002 or 2003, I think) we have generally moved
>>> towards trying to avoid using it. Ipv4 works pretty much everywhere,
>>> ipv6 still works almost nowhere.
>>
>> Welcome to the real world:
>>
>> http://www.inetcore.com/project/ipv4ec/index_en.html
>>
>> It's not a question of "ipv4 works pretty much everywhere, ipv6 still
>> works almost nowhere". Maybe that was true a few yaers ago, but nowadays
>> the ipv6 network is getting more and more mature. Avoiding ipv6 in 2009
>> is ostrich policy.
>
> Maybe. But we're not avoiding ipv6 just to avoid it. We've just been
> turning off ipv6 features in response to bug reports that shows us cases
> where using ipv6 would make opera fail completely. As long as ipv4
> actually works, it is better to successfully use ipv4 than to fail to
> use ipv6. Of course, if ipv4 no longer works everywhere, then we will
> have to live with failing completely in some cases. We just have to
> pick our poison.
That is a pretty pathetic response. What you are saying is that if it's
broken, you'll remove it rather than fix it. If you did with the rest of
Opera, you'll make those who are complaining about bloat very happy as the
size will be reduced to zero!!!
B
Eirik Byrkjeflot Anonsen
Not at all. We definitely prefer fixing code, but that is not always an
option. Given the choice of shipping a broken product and a working
product, we prefer shipping a working product.
Some of the problems we have had has been near impossible to fix. What
do you do when the system insists that ipv6 is the right thing to use,
but it doesn't work?
I'm sure the people working on this are trying to fix it. I assume
fixes have been made. But I also assume there are still plenty of cases
where it is hard to tell if trying to use ipv6 will be a bad idea. If
we have to choose between ipv4 and ipv6, we want to choose the one more
likely to work. If we have clear indications that ipv6 will work, (or
that ipv4 will not work) we should try that. But lacking such
information (or if the information is known to be untrustworthy), I
still believe that in today's internet, ipv4 is orders of magnitude more
likely to work than ipv6.
Rest assured that we do want to use ipv6 when available. We just have
to figure out how to do it safely.
eirik
Christof Meerwald
[...]
> Some of the problems we have had has been near impossible to fix. What
> do you do when the system insists that ipv6 is the right thing to use,
> but it doesn't work?
An easy workaround would be to have a configuration option to let the user
enable IPv6 support (and it could even be disabled by default).
But why is it that other browsers seem to be able to cope with the
situation?
Tore Anderson
> But why is it that other browsers seem to be able to cope with the
> situation?
All other major browsers support RFC 3484. Opera doesn't.
For more information, see:
http://my.opera.com/community/forums/topic.dml?id=298893
http://thread.gmane.org/gmane.org.operators.ipv6/2636
http://thread.gmane.org/gmane.org.operators.ipv6/2683
Best regards,
--
Tore Anderson
Eirik Byrkjeflot Anonsen
> On Fri, 04 Dec 2009 16:21:55 +0100, Eirik Byrkjeflot Anonsen wrote:
> [...]
>> Some of the problems we have had has been near impossible to fix. What
>> do you do when the system insists that ipv6 is the right thing to use,
>> but it doesn't work?
>
> An easy workaround would be to have a configuration option to let the user
> enable IPv6 support (and it could even be disabled by default).
>
> But why is it that other browsers seem to be able to cope with the
> situation?
I'm not saying we have done a good job out of ipv6. I find it quite
likely that we have done a pretty bad job out of ipv6. Again, big
disclaimer: I don't know much about what we have done with ipv6, as it
is nowhere near what I'm working on. That being said, I can think of a
couple of possible reasons why our ipv6 support may be bad:
- It is not perceived as an important issue by the people who prioritize
issues nor by the people directly working on them. After all, opera
works on most of the internet. We have plenty of other issues that
really breaks opera, and those may just be perceived as more important
to fix.
- None of the people involved actually are on a primarily-ipv6 system.
Thus there is no "itch to scratch". (Unlike, for example, the amd64
issues.)
I don't know if the above is true, but I think they are at least
reasonable suggestions.
eirik
Bastiaan Welmers
> Just noticed that with Opera 10.10 (on Ubuntu 9.10 AMD64), IPv6 is
now
> completely broken - previously it was only half broken (it didn't
fallback
> from IPv6 to IPv4 if there was any problem), but now it doesn't even
try
> IPv6 any more - not even if the site is IPv6 only.
>
Recently upgraded to Opera 10.10 and found the same problematic
behaviour. Since our intranet is only browsable via ipv6 Opera
is not useable anymore here. Sad this step had taken.
/Bastiaan
Damyan Yordanov
Sad but true. Still waitin' for a fix…
Damyan
--
Using Debian Testing (Squeeze), Kernel 2.6.22-3-686, Postfix 2.6.5-3, Dovecot 1:1.2.9-1, Cone 0.79-2, Tin 1:1.9.5
Freeaqingme
upgrade to opera 10.10 just made it impossible for me to do my work
using opera (and that while I like the improvements done to
dragonfly). Now having to resort to inferior browsers like FF (which
actually does ipv6 flawlessly). Now I can imagine the bugs being a
pain in the ass if solving them involve really big architectural
changes (I'm a developer myself, lol), but make it optional then to
enable ipv6, so people who rely on ipv6 can at least turn it on (or
better; turn it on by default, and let it be turned off by people
having problems with it). I'm aware that not that many people are
dependent on ipv6, but please note that that is exactly because of
problems like these; everybody is just too lazy to decently implement
it, and just points fingers at others, saying that "others" should
implement it first. By the way, ever thought of what we're going to do
when there are no more ipv4 addresses available? http://www.ipv4countdown.com/
Regards,
Dolf Schimmel
-- Freeaqingme
p.s. Am I right in saying the voice controlls were also removed as of
10.x?
Jorgen Grahn
> Being a webdeveloper with an ipv6-only development environment an
> upgrade to opera 10.10 just made it impossible for me to do my work
> using opera (and that while I like the improvements done to
> dragonfly). Now having to resort to inferior browsers like FF (which
> actually does ipv6 flawlessly).
Or the "inferior" release of Opera which came before 10.10.
I'm a fan of IPv6 too and find this anti-IPv6 decision disturbing, but
there's really no need to be that melodramatic.
/Jorgen
--
// Jorgen Grahn <grahn@ Oo o. . .
\X/ snipabacken.se> O o .
Eirik Byrkjeflot Anonsen
> On Sat, 2010-01-30, Freeaqingme wrote:
>> Being a webdeveloper with an ipv6-only development environment an
>> upgrade to opera 10.10 just made it impossible for me to do my work
>> using opera (and that while I like the improvements done to
>> dragonfly). Now having to resort to inferior browsers like FF (which
>> actually does ipv6 flawlessly).
>
> Or the "inferior" release of Opera which came before 10.10.
>
> I'm a fan of IPv6 too and find this anti-IPv6 decision disturbing, but
> there's really no need to be that melodramatic.
On the good side, now that IPv6 in opera finally decided to break down
completely, I hear we're working on fixing it "properly". (It does help
that we suddenly discovered that there is an RFC on this. Pretty damn
stupid of us not to have thought of that ourselves...)
(I don't know much about what they're doing to fix it, or whether there
are architectural issues. Or even how it is prioritized. So I can't
say anything about when this fix will be done.)
eirik
Tommy A. Olsen
> Jorgen Grahn <grahn+n...@snipabacken.se> writes:
> > On Sat, 2010-01-30, Freeaqingme wrote:
> >> Being a webdeveloper with an ipv6-only development environment an
> >> upgrade to opera 10.10 just made it impossible for me to do my work
> >> using opera (and that while I like the improvements done to
> >> dragonfly). Now having to resort to inferior browsers like FF (which
> >> actually does ipv6 flawlessly).
>
> > Or the "inferior" release of Opera which came before 10.10.
>
> > I'm a fan of IPv6 too and find this anti-IPv6 decision disturbing, but
> > there's really no need to be that melodramatic.
>
> On the good side, now that IPv6 in opera finally decided to break down
> completely, I hear we're working on fixing it "properly". (It does help
> that we suddenly discovered that there is an RFC on this. Pretty damn
> stupid of us not to have thought of that ourselves...)
Actually, Opera 10.10 has a silly bug because we intended to fix one
of the aspects of IPv6 that weren't working. It lead to Opera turning
off IPv6 lookup alltogether if the interface had any kind of IPv6
address, even an fe80 (link-locl) one. This is fixed in the next
snapshot of 10.50 for Unix (also internally on the peregrine branch),
where we've also added support for Teredo and 6to4, where support
means to prefer IPv4 if this is the kind of IPv6 interface you have.
If you have native IPv6, it prefers that.
Regarding the RFC, we currently only support that fully on Windows,
where the solution was as simple as trusting getaddrinfo() and its
sorting order, which complies with the RFC. We hope to get there on
UNIX and mac as well in the near future. For now, the measures we've
taken will work for 99.9% of people.
Jorgen Grahn
> On Feb 2, 9:08�am, Eirik Byrkjeflot Anonsen <ei...@opera.com> wrote:
>> On the good side, now that IPv6 in opera finally decided to break down
>> completely, I hear we're working on fixing it "properly". �(It does help
>> that we suddenly discovered that there is an RFC on this. �Pretty damn
>> stupid of us not to have thought of that ourselves...)
> Regarding the RFC, we currently only support that fully on Windows,
> where the solution was as simple as trusting getaddrinfo() and its
> sorting order, which complies with the RFC. We hope to get there on
> UNIX and mac as well in the near future. For now, the measures we've
> taken will work for 99.9% of people.
Completely unrelated to Opera, but are you saying there's a bug/RFC
non-compliance in getaddrinfo() on Unix, or that you currently cannot
use it properly?
Tommy A. Olsen
> On Sat, 2010-02-06, Tommy A. Olsen wrote:
> Completely unrelated to Opera, but are you saying there's a bug/RFC
> non-compliance in getaddrinfo() on Unix, or that you currently cannot
> use it properly?
Currently we can't use it properly on our Unix platforms because of
complications with our current implementation. Lack of time is the
dominating factor here really, and we really want to implement it the
right way.
Eirik Byrkjeflot Anonsen
> On Sat, 2010-02-06, Tommy A. Olsen wrote:
>> On Feb 2, 9:08 am, Eirik Byrkjeflot Anonsen <ei...@opera.com> wrote:
> ...
>
> Completely unrelated to Opera, but are you saying there's a bug/RFC
> non-compliance in getaddrinfo() on Unix, or that you currently cannot
> use it properly?
I remember Morten discovering getaddrinfo() and he was very happy to be
able to throw out all of our clunky, ifdeffed gethostbyname() code. And
I remember him swearing loudly a few days later when he found that the
implementation of getaddrinfo() on various platforms was just as
inconsistent as gethostbyname().
Or in other words: what do you mean by "unix"?
(Things may have improved since then, of course. I think this was in
2002. Note also that Tommy says that on windows getaddrinfo() gives us
the addresses in preferred order. I haven't read the getaddrinfo()
specification(s), but it may well be that this is not a required
behaviour for getaddrinfo().)
eirik
Jorgen Grahn
Interesting. Stevens' "Unix Network Programming" (1998) pushed hard
for getaddrinfo(), but if I recall correctly it was pretty new back
then, and he himself was involved in its design (V4/V6-independence
and so on), so I kind of suspected it didn't turn out *that* well in
practice.
(Not that I'm doing a lot of IPv6-enabled, cross-platform network
programming or anything. I use getaddrinfo() because it's so
convenient, but my target is Linux machines in all-IPv4 networks.)
Christof Meerwald
[...]
> able to throw out all of our clunky, ifdeffed gethostbyname() code. And
> I remember him swearing loudly a few days later when he found that the
> implementation of getaddrinfo() on various platforms was just as
> inconsistent as gethostbyname().
BTW, IPv6 appears to work again for me with the latest Opera snapshots.
Chris Hills
<NOSPAM-see...@usenet.cmeerw.org> wrote:
> Just noticed that with Opera 10.10 (on Ubuntu 9.10 AMD64), IPv6 is now
> completely broken - previously it was only half broken (it didn't
> fallback
> from IPv6 to IPv4 if there was any problem), but now it doesn't even try
> IPv6 any more - not even if the site is IPv6 only.
With 10.50-6240 on Linux IPv6 is working again (and probably has been for
a few releases, but this is the latest that I am using right now).
Christof Meerwald
> With 10.50-6240 on Linux IPv6 is working again (and probably has been for
> a few releases, but this is the latest that I am using right now).
Unfortunately, IPv6 support is still very limited - i.e. it doesn't fall
back to IPv4 if it isn't able to connect via IPv6.
To illustrate the issue I have now set up
http://edge.cmeerw.net/ipv4-fallback.html
If the browser properly supports IPv6 and you have IPv6 connectivity you
should see:
Your browser managed to load the IPv4 style sheet.
You seem to have IPv6 connectivity.
However, in Opera you currently see (if you have IPv6 connectivity):
Your browser didn't load the IPv4 style sheet.
You seem to have IPv6 connectivity.
Christof Meerwald
> Unfortunately, IPv6 support is still very limited - i.e. it doesn't fall
> back to IPv4 if it isn't able to connect via IPv6.
>
> To illustrate the issue I have now set up
> http://edge.cmeerw.net/ipv4-fallback.html
This is also still broken in the final 10.60 build.