ipv6 and stock Arch kernel26

pixel · 2006-04-04 00:10:41

http://bugs.archlinux.org/task/4357

Please vote, I know most Arch users don't need ipv6 so it would be only fair to make it as an add-on module only in the stock kernel26.
Right now it's impossible to turn off ipv6 as it's built in inside kernel.

cascat · 2006-04-04 02:40:07

May I suggest using a module in -beyond kernel as well

JGC · 2006-04-04 09:31:20

modularized ipv6 is the way to go, most users don't use it, but some might want it. For the -beyond and -archck kernels: it's logical to have the feature sets the same between stock kernels and official tweaked stock kernels. If we make something a module in the stock kernel, it should be a module in these too.

iphitus · 2006-04-04 09:32:42

JGC wrote:

modularized ipv6 is the way to go, most users don't use it, but some might want it. For the -beyond and -archck kernels: it's logical to have the feature sets the same between stock kernels and official tweaked stock kernels. If we make something a module in the stock kernel, it should be a module in these too.

That's what I do, I tend to mirror my config to match the vanilla one.

Neuro · 2006-04-04 12:26:07

JGC wrote:

modularized ipv6 is the way to go, most users don't use it, but some might want it. For the -beyond and -archck kernels: it's logical to have the feature sets the same between stock kernels and official tweaked stock kernels. If we make something a module in the stock kernel, it should be a module in these too.

can't agree more

test1000 · 2006-04-05 00:16:19

just out of curiousity; if one setup the ipv6 kernel module to also control ipv4, would it do it faster than the ipv4 module?

i've heard it's better/faster code..

shadowhand · 2006-04-05 00:22:40

NO!

The reason that IPv6 has such poor adoption is that so many people refuse to support it. By moving IPv6 to a module, a LOT of people are going to stop using it, thus aren't going to be supporting it.

IPv6 very very seldom causes issues with anyone's network, but if it does, there are hacks to disable it from taking effect. Please show support for IPv6 and keep it built into the kernel!

iphitus · 2006-04-05 00:51:11

It's stupid fixing a userspace problem with a kernel space hack.

For that reason alone, ipv6 has no reason to be modularised.

And the aforementioned bug ought to be on firefox's bugzilla, not our flyspray.

James

cactus · 2006-04-05 01:34:45

shadowhand wrote:

NO!
The reason that IPv6 has such poor adoption is that so many people refuse to support it. By moving IPv6 to a module, a LOT of people are going to stop using it, thus aren't going to be supporting it.
IPv6 very very seldom causes issues with anyone's network, but if it does, there are hacks to disable it from taking effect. Please show support for IPv6 and keep it built into the kernel!

ipv6 is also not supported by hardly any upstream providers, and the hassle with ipv6 to ipv4 conversion is a pain in the butt.
The only place it is currently usable is for internal networks. Until upstream providers start making use of it (or people start tunneling ipv6 over ipv4..blech..) then it is generally useless to have it enabled.

shadowhand · 2006-04-05 02:01:18

Upstream providers aren't using it because their clients don't use it. How's that for an infinate loop?

while( count($users_with_ipv6) < $really_big_number )
{
  echo "IPv6: YOU DON'T GOT NO ADOPTION!";
  sleep(5);
}

echo "Oh crap, we have to support it now... :("

Snarkout · 2006-04-05 03:41:38

shadowhand wrote:

Upstream providers aren't using it because their clients don't use it. How's that for an infinate loop?
while( count($users_with_ipv6) < $really_big_number )
{
  echo "IPv6: YOU DON'T GOT NO ADOPTION!";
  sleep(5);
}

echo "Oh crap, we have to support it now... :("

It's extremely unlikely that end-users will dictate when ipv6 becomes reality.

xterminus · 2006-04-05 04:08:12

cactus wrote:

ipv6 is also not supported by hardly any upstream providers, and the hassle with ipv6 to ipv4 conversion is a pain in the butt.
The only place it is currently usable is for internal networks. Until upstream providers start making use of it (or people start tunneling ipv6 over ipv4..blech..) then it is generally useless to have it enabled.

Ipv6 is trivial to get going. Esp. if you use 6to4 tunneling, which builds a tunnel automatically to the closest ip6 router to you.

Slap the following into rc.local and you should be set to go.

# Figure out current ip address on eth0
ipv4addr=$(ip addr show dev eth0 | grep "inet " | sed -e 's/s*inet ([0-9]*.[0-9]*.[0-9]*.[0-9]*).*/1/')
# Create a ipv6 address based on ipv4 address
ipv6addr=$(printf "2002:%02x%02x:%02x%02x::1" ${ipv4addr//./ })
# bring up tunnel
ip link set dev tun6to4 up &&
# assign tunnel an ip address
ip -6 addr add $ipv6addr/16 dev tun6to4 &&
# add a default route to nearest ip6 router
ip -6 route add 2000::/3 via ::192.88.99.1 dev tun6to4 metric 1

brain0 · 2006-04-05 05:47:40

shadowhand wrote:

Upstream providers aren't using it because their clients don't use it. How's that for an infinate loop?

And you want to show them that you "use" ipv6 by having it enabled? That's nonsense.

xterminus wrote:

Ipv6 is trivial to get going. Esp. if you use 6to4 tunneling, which builds a tunnel automatically to the closest ip6 router to you.

When I used a 6to4 anycast gateway, many websites that also had 2001:: addresses were not accessible any more (I had timeouts) or only with very high latency. I switched it off again for that reason.
Reaching other 6to4 users with 2002:: addresses worked well - although I only knew one other user at that time.

Neuro · 2006-04-05 06:19:59

shadowhand wrote:

Upstream providers aren't using it because their clients don't use it. How's that for an infinate loop?

ipv6 would be great if... it were used broadly. I mean, what's the point? Unless the big ISPs start serving their normal customers (by normal I mean their corporate clients and even average Joe), ipv6 won't take off.

Don't get me wrong, it's really cool in, for example, university networks. At least that's what the netadmin of my (hopefully) future uni says.

But tunneling ipv6 through ipv4 is marely a hack. What does it serve, what's the purspose? It ain't faster, since you still have to use the ipv4 protocol and it might cause problems just because of it's complexity. Or am I missing something?

codemac · 2006-04-05 21:43:32

I voted yes because I thought I was voting for ipv6.

Hmm.

Well, I think ipv6 should be default. End users will not dictate when ipv6 becomes the standard, but it all ready is a reality. It's better. Period. As a knowledgable end user, you should try to use it.

WillySilly · 2006-04-05 21:47:28

I concur with codemac

cactus · 2006-04-05 22:41:31

Yet another reason to keep compiling my own kernels.
*sigh*

iphitus · 2006-04-06 00:09:45

cactus wrote:

Yet another reason to keep compiling my own kernels.
*sigh*

What are the others?

mcubed · 2006-04-06 01:02:28

I sincerely hope IPv6 will be modularized. It has given me nothing but grief for over three months on both Debian and Ubuntu. If it starts giving me trouble on Arch, I'll have to find another distro again (sigh), or give up on Linux altogether and go back to OS X. Or maybe FreeBSD. I've had no connection issues on OS X, which uses Darwin, so it's possible that the BSDs would be bug-free vis-a-vis IPv6. Linux certainly hasn't been.

But really, what's the point? No one uses IPv6 because it's a solution in search of a problem. It's poorly executed and poorly managed. I hope it dies a painful death.

Neuro · 2006-04-06 06:23:54

codemac wrote:

Well, I think ipv6 should be default. End users will not dictate when ipv6 becomes the standard, but it all ready is a reality. It's better. Period. As a knowledgable end user, you should try to use it.

If the end users will not dictate when ipv6 becomes the standard, what's the point in persuading them to use it if it's useless? I can't think of any reason, any field in which the ipv6 over ipv4 tunneling is better than the usual ipv4. I agree, it is better if it's use as a original backend, but since pretty much no one has a ipv6 based connection at home, why bother? Just provide one valid reason for using some possiblt unstable hack-like ipv6 over ipv4 tunneling instead of a normal ipv4 infrastructure. Apart from the nerd factor of using the most bleeding edge stuff..

mcubed wrote:

But really, what's the point? No one uses IPv6 because it's a solution in search of a problem. It's poorly executed and poorly managed. I hope it dies a painful death.

On the other hand, I think your statement is a bit exaggerated. It is better in many ways, starting with the elimination of NAT due to immense address space. The simple fact is: ipv6 is better than ipv4. It's that ipv6 tunneling over ipv4 is not.

iphitus · 2006-04-06 06:37:11

mcubed wrote:

I sincerely hope IPv6 will be modularized. It has given me nothing but grief for over three months on both Debian and Ubuntu. If it starts giving me trouble on Arch, I'll have to find another distro again (sigh), or give up on Linux altogether and go back to OS X. Or maybe FreeBSD. I've had no connection issues on OS X, which uses Darwin, so it's possible that the BSDs would be bug-free vis-a-vis IPv6. Linux certainly hasn't been.
But really, what's the point? No one uses IPv6 because it's a solution in search of a problem. It's poorly executed and poorly managed. I hope it dies a painful death.

If it's giving you so much problem on Arch, then why havn't you posted for help?

The only problem I've seen it cause on arch has been firefox, and thats for the firefox devs to fix, not for us. If there's any other problems it causes, I'd like to hear them, because unless I do - kernel26beyond will be keeping ipv6 as it is now and kernel26 probably will too.

James

mcubed · 2006-04-06 10:09:44

Neuro wrote:

It is better in many ways, starting with the elimination of NAT due to immense address space. The simple fact is: ipv6 is better than ipv4. It's that ipv6 tunneling over ipv4 is not.

I never said it was worse than IPv4, I said it's being poorly managed. The theory is better, sure, but the reality is a joke. People can come up with lots of great ideas; but implementing them is another story. You have to question whether the hassle is worth the perceived gains. If there was a great unmet demand that IPv6 could satisfy or great potential that IPv6 could offer that isn't available now, then the roll-out would've happened by now. Take a look around. Hardly anyone is using it. Doesn't that tell you something?

iphitus wrote:

If it's giving you so much problem on Arch, then why havn't you posted for help?

It has not caused be problems on Arch, not yet. It would be wonderful if whatever was causing it has been fixed. But I don't trust it, not after all the headaches on Debian and Ubuntu, and not just with Firefox. Until I disabled it, I couldn't get anywhere except with Konqueror or w3m -4 (the option to use IPv4 only; if I didn't specify -4, w3m wouldn't even load Google). Even after disabling IPv6, I couldn't connect to the apt servers without pinging them first, nor could I use Bittorrent without pinging. No protocol was fully functional. Disabling IPv6 at least let me browse the web without too much hassle.

This goes way beyond Firefox, or any other Gecko browser. Obviously not everyone has these problems, but you should be aware that it has created big problems for some people, AFAIK only Linux users. I've not heard of these issues with OS X, the BSDs or Windows. I was dual booting Debian or Ubuntu with OS X on the same iMac, connected to the same DSL modem provided by the same ISP, and Firefox (and Camino) worked just fine under OS X, as did everything else. It seems strange to me that Linux developers are trying to palm off their own screw-ups on a x-platform browser that works perfectly well under any OS except Linux. When I was searching for a fix, I found a thread on the Debian dev list dating from 2004 where people were complaining about these problems, and the devs were trying to lay it on Mozilla.

Last time I checked, w3m doesn't use Gecko to render its text!

iphitus · 2006-04-06 10:59:50

mcubed wrote:

Neuro wrote:
It is better in many ways, starting with the elimination of NAT due to immense address space. The simple fact is: ipv6 is better than ipv4. It's that ipv6 tunneling over ipv4 is not.
I never said it was worse than IPv4, I said it's being poorly managed. The theory is better, sure, but the reality is a joke. People can come up with lots of great ideas; but implementing them is another story. You have to question whether the hassle is worth the perceived gains. If there was a great unmet demand that IPv6 could satisfy or great potential that IPv6 could offer that isn't available now, then the roll-out would've happened by now. Take a look around. Hardly anyone is using it. Doesn't that tell you something?

But the number of people using it grow, every once in a while I do see a site that offers an ipv6 version.

My old computer came with USB and a DVD drive. This computer was a 300mhz P2. Neither of these technologies were used extensively for many many years to come, but when DVDs and USB came into widespread use, I was damned happy I had them there.

Because it isnt widespread at all now, doesnt mean it shouldnt be included.

iphitus wrote:
If it's giving you so much problem on Arch, then why havn't you posted for help?
It has not caused be problems on Arch, not yet. It would be wonderful if whatever was causing it has been fixed. But I don't trust it, not after all the headaches on Debian and Ubuntu, and not just with Firefox. Until I disabled it, I couldn't get anywhere except with Konqueror or w3m -4 (the option to use IPv4 only; if I didn't specify -4, w3m wouldn't even load Google). Even after disabling IPv6, I couldn't connect to the apt servers without pinging them first, nor could I use Bittorrent without pinging. No protocol was fully functional. Disabling IPv6 at least let me browse the web without too much hassle.

Sounds like problems in your configuration. Not the fault of the kernel.

The kernel provides an interface. It's up to the applications to choose whether or not they want to use it.

When you get problems like above, the interface has been misused or there is a misconfiguration somewhere.

When I was searching for a fix, I found a thread on the Debian dev list dating from 2004 where people were complaining about these problems, and the devs were trying to lay it on Mozilla.
Last time I checked, w3m doesn't use Gecko to render its text!

And they were rightly laying it on Mozilla. If it causes the slowdowns and problems that everyone claims it does, then that is a problem with firefox. If firefox is choosing to use ipv6, then thats a fault in firefox, not the kernel. Same goes for w3m.

Although the problem you describe, as it affects more than one program, sounds more like a global configuration issue or debian userspace fault. If it were a specific bug in the browser, w3m, then that bug would exist on Arch too. It doesnt.

Userspace bugs should not be fixed in the kernel.

James

pixel · 2006-04-06 11:01:58

For me it has really NOTHING to do with Firefox or other userspace application. ipv6 should be modularized because its adoption is not widespread. I would say that most people don't use it and hence don't need it. Why we need to bloat the kernel with something we won't really use?

iphitus · 2006-04-06 11:08:45

pixel wrote:

For me it has really NOTHING to do with Firefox or other userspace application. ipv6 should be modularized because its adoption is not widespread. I would say that most people don't use it and hence don't need it. Why we need to bloat the kernel with something we won't really use?

pixel wrote:

Judd,
http://bbs.archlinux.org/viewtopic.php? … sable+ipv6
http://bbs.archlinux.org/viewtopic.php? … sable+ipv6
http://bbs.archlinux.org/viewtopic.php? … sable+ipv6
It seems there is also problem with the speed of Firefox and ipv6, more threads are about this.
Dale, don't be ironic.. how many people really use ipv6? ipv4 is essential on the other hand....
If you want to know the opinion of the community on this make a poll, im sure most would like to see ipv6 as a module.

It has everything to do with it according to you

James

Arch Linux

#1 2006-04-04 00:10:41

ipv6 and stock Arch kernel26

#2 2006-04-04 02:40:07

Re: ipv6 and stock Arch kernel26

#3 2006-04-04 09:31:20

Re: ipv6 and stock Arch kernel26

#4 2006-04-04 09:32:42

Re: ipv6 and stock Arch kernel26

#5 2006-04-04 12:26:07

Re: ipv6 and stock Arch kernel26

#6 2006-04-05 00:16:19

Re: ipv6 and stock Arch kernel26

#7 2006-04-05 00:22:40

Re: ipv6 and stock Arch kernel26

#8 2006-04-05 00:51:11

Re: ipv6 and stock Arch kernel26

#9 2006-04-05 01:34:45

Re: ipv6 and stock Arch kernel26

#10 2006-04-05 02:01:18

Re: ipv6 and stock Arch kernel26

#11 2006-04-05 03:41:38

Re: ipv6 and stock Arch kernel26

#12 2006-04-05 04:08:12

Re: ipv6 and stock Arch kernel26

#13 2006-04-05 05:47:40

Re: ipv6 and stock Arch kernel26

#14 2006-04-05 06:19:59

Re: ipv6 and stock Arch kernel26

#15 2006-04-05 21:43:32

Re: ipv6 and stock Arch kernel26

#16 2006-04-05 21:47:28

Re: ipv6 and stock Arch kernel26

#17 2006-04-05 22:41:31

Re: ipv6 and stock Arch kernel26

#18 2006-04-06 00:09:45

Re: ipv6 and stock Arch kernel26

#19 2006-04-06 01:02:28

Re: ipv6 and stock Arch kernel26

#20 2006-04-06 06:23:54

Re: ipv6 and stock Arch kernel26

#21 2006-04-06 06:37:11

Re: ipv6 and stock Arch kernel26

#22 2006-04-06 10:09:44

Re: ipv6 and stock Arch kernel26

#23 2006-04-06 10:59:50

Re: ipv6 and stock Arch kernel26

#24 2006-04-06 11:01:58

Re: ipv6 and stock Arch kernel26

#25 2006-04-06 11:08:45

Re: ipv6 and stock Arch kernel26

Board footer