Hey, folks. If you follow the trac ticket notifications on test@ you may have noticed this already, but I know it's easy to tune those out, and I'm CCing devel@ so people following that list know about this too.
We're planning a Fedora IPv6 Test Day to coincide with World IPv6 Day on June 8. This has been kicked off by some folks inside Red Hat whose job it is to understand IPv6 (those poor, poor kids), so thanks to Linda and her team for bringing it to the Fedora community. We have a trac ticket for co-ordinating the event here:
https://fedorahosted.org/fedora-qa/ticket/192
and the Wiki page for it is here:
https://fedoraproject.org/wiki/QA/Test_Day:2011-06-08_IPv6
it would be really good if people can pencil in this event on their calendars so we can get some testing of Fedora 15's IPv6 preparedness, and it would also be great if those who understand what the hell they're doing with IPv6 could check over the Wiki page for accuracy and completeness. Finally it would also help if people who don't understand what the hell they're doing with IPv6 - like me! - could try and follow the instructions, and let us know - via this thread, or the trac ticket - if they struggle or find something unclear or don't seem to be able to make it work as intended, so we can improve the instructions. I've already noted, for example, that it's not particularly clear what you should do if you're using a router - whether to try and set the tunnel up on the router, or whether you can just do it on one of the machines behind the router.
Thanks everyone!
On Tue, May 31, 2011 at 8:28 PM, Adam Williamson awilliam@redhat.com wrote:
it would be really good if people can pencil in this event on their calendars so we can get some testing of Fedora 15's IPv6 preparedness, and it would also be great if those who understand what the hell they're doing with IPv6 could check over the Wiki page for accuracy and completeness.
Is there an IPv6 tracker bug? There is for instance an issue with iptables filtering and DHCPv6 and maybe more of those:
RHBZ #591630 DHCPv6 responses are not allowed by default ip6tables ruleset RHBZ #552099 system-config-firewall has no simple mechanism to enable IPv6 DHCPv6 client RHBZ #656334 Default Firewall blocking DHCPv6
(They all seem to be about the same problem...)
Finally it would also help if people who don't understand what the hell they're doing with IPv6 - like me! - could try and follow the instructions, and let us know - via this thread, or the trac ticket
This [1] may be of some help as a high level overview of how to deploy IPv6 on a LAN and various operating system IPv6 compatibilities. Fedora is doing quite well! The document is not a configuration help, but it might make it clear how everything fits together and brush you up on your IPv6 :-)
Regards, François
[1] http://www.surfnet.nl/Documents/IPv6%20Deployment%20In%20Local%20Area%20Netw...
On Wed, Jun 01, 2011 at 08:22:25AM +0200, fkooman@tuxed.net wrote:
This [1] may be of some help as a high level overview of how to deploy IPv6 on a LAN and various operating system IPv6 compatibilities. Fedora is doing quite well! The document is not a configuration help, but it might make it clear how everything fits together and brush you up on your IPv6 :-)
[...]
[1] http://www.surfnet.nl/Documents/IPv6%20Deployment%20In%20Local%20Area%20Netw...
This document is a good introduction, but it appears to assume that I've got an IPv6 router and internet connection. I guess that's not going to apply to very many people.
Is there an easy way I can set up IPv6 and a handful of machines on my LAN for testing, without requiring any IPv6 internet connection or an IPv6 assigned prefix?
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000 link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft forever
but pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
(Note I have no idea at all what I'm doing here so it's probably some elementary error).
Rich.
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
On Wed, Jun 01, 2011 at 08:22:25AM +0200, fkooman@tuxed.net wrote:
This [1] may be of some help as a high level overview of how to deploy IPv6 on a LAN and various operating system IPv6 compatibilities. Fedora is doing quite well! The document is not a configuration help, but it might make it clear how everything fits together and brush you up on your IPv6 :-)
[...]
[1] http://www.surfnet.nl/Documents/IPv6%20Deployment%20In%20Local%20Area%20Netw...
This document is a good introduction, but it appears to assume that I've got an IPv6 router and internet connection. I guess that's not going to apply to very many people.
Is there an easy way I can set up IPv6 and a handful of machines on my LAN for testing, without requiring any IPv6 internet connection or an IPv6 assigned prefix?
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000 link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft forever
but pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this. The only thing that commonly uses the link local addresses is IPv6 autoconfig, at which point you get a real IPv6 address, which is globally unique that real applications can use.
Daniel
Daniel P. Berrange wrote:
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft foreverbut pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this.
Ping6 has a way though:
$ ping6 -I eth0 fe80::200:24ff:fec9:2e0c PING fe80::200:24ff:fec9:2e0c(fe80::200:24ff:fec9:2e0c) from fe80::21e:8cff:fecf:cde5 eth0: 56 data bytes 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=1 ttl=64 time=1.69 ms 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=2 ttl=64 time=0.263 ms
Björn Persson
On Thu, Jun 02, 2011 at 06:00:44PM +0200, Björn Persson wrote:
Daniel P. Berrange wrote:
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft foreverbut pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this.
Ping6 has a way though:
$ ping6 -I eth0 fe80::200:24ff:fec9:2e0c PING fe80::200:24ff:fec9:2e0c(fe80::200:24ff:fec9:2e0c) from fe80::21e:8cff:fecf:cde5 eth0: 56 data bytes 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=1 ttl=64 time=1.69 ms 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=2 ttl=64 time=0.263 ms
But that's not so useful if no other programs can use these addresses. Is there a way to assign "proper" addresses? Just run radvd and configure it with a random prefix?
Rich.
On Thu, 2 Jun 2011 17:07:47 +0100 "Richard W.M. Jones" rjones@redhat.com wrote:
On Thu, Jun 02, 2011 at 06:00:44PM +0200, Björn Persson wrote:
Daniel P. Berrange wrote:
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft foreverbut pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this.
Ping6 has a way though:
$ ping6 -I eth0 fe80::200:24ff:fec9:2e0c PING fe80::200:24ff:fec9:2e0c(fe80::200:24ff:fec9:2e0c) from fe80::21e:8cff:fecf:cde5 eth0: 56 data bytes 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=1 ttl=64 time=1.69 ms 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=2 ttl=64 time=0.263 ms
But that's not so useful if no other programs can use these addresses. Is there a way to assign "proper" addresses? Just run radvd and configure it with a random prefix?
Instead of random, you may want to use a Unique Local Address prefix, following http://www.ripe.net/lir-services/resource-management/ipv6/ipv6-address-types
Rich.
On Thu, Jun 02, 2011 at 12:16:13PM -0400, Bernd Stramm wrote:
On Thu, 2 Jun 2011 17:07:47 +0100 "Richard W.M. Jones" rjones@redhat.com wrote:
On Thu, Jun 02, 2011 at 06:00:44PM +0200, Björn Persson wrote:
Daniel P. Berrange wrote:
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft foreverbut pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this.
Ping6 has a way though:
$ ping6 -I eth0 fe80::200:24ff:fec9:2e0c PING fe80::200:24ff:fec9:2e0c(fe80::200:24ff:fec9:2e0c) from fe80::21e:8cff:fecf:cde5 eth0: 56 data bytes 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=1 ttl=64 time=1.69 ms 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=2 ttl=64 time=0.263 ms
But that's not so useful if no other programs can use these addresses. Is there a way to assign "proper" addresses? Just run radvd and configure it with a random prefix?
Instead of random, you may want to use a Unique Local Address prefix, following http://www.ripe.net/lir-services/resource-management/ipv6/ipv6-address-types
It works :-)
https://rwmj.wordpress.com/2011/06/02/ipv6-lan/
Rich.
Am Donnerstag, den 02.06.2011, 12:16 -0400 schrieb Bernd Stramm:
On Thu, 2 Jun 2011 17:07:47 +0100 "Richard W.M. Jones" rjones@redhat.com wrote:
On Thu, Jun 02, 2011 at 06:00:44PM +0200, Björn Persson wrote:
Daniel P. Berrange wrote:
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000
link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft foreverbut pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this.
Ping6 has a way though:
$ ping6 -I eth0 fe80::200:24ff:fec9:2e0c PING fe80::200:24ff:fec9:2e0c(fe80::200:24ff:fec9:2e0c) from fe80::21e:8cff:fecf:cde5 eth0: 56 data bytes 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=1 ttl=64 time=1.69 ms 64 bytes from fe80::200:24ff:fec9:2e0c: icmp_seq=2 ttl=64 time=0.263 ms
But that's not so useful if no other programs can use these addresses. Is there a way to assign "proper" addresses? Just run radvd and configure it with a random prefix?
Instead of random, you may want to use a Unique Local Address prefix, following http://www.ripe.net/lir-services/resource-management/ipv6/ipv6-address-types
I brought this up some time ago: Assigning ULA instead of LL would be somewhat more user friendly: https://bugzilla.gnome.org/show_bug.cgi?id=615063 https://bugzilla.redhat.com/show_bug.cgi?id=581579
In my eyes just assigning a LL adress is not very user friendly.
Rich.
-- Bernd Stramm bernd.stramm@gmail.com
On Thu, 2011-06-02 at 16:48 +0100, Daniel P. Berrange wrote:
Anything with an fe80:: prefix is a link local address, which is only unique within the scope of a single LAN segment. Thus if you want to send traffic to such addresses, you need to specify the NIC to send the traffic out from. The vast majority of apps using sockets have no way to let you do this.
[dwmw2@i7 activesyncd]$ ssh fe80::21d:7dff:fe04:dbe2%eth0 Last login: Tue Jun 21 12:40:41 2011 from i7.infradead.org [dwmw2@twosheds ~]$
wget fails though: [dwmw2@i7 activesyncd]$ wget 'http://%5Bfe80::21d:7dff:fe04:dbe2%eth0%5D/' http://%5Bfe80::21d:7dff:fe04:dbe2%eth0%5D/: Invalid IPv6 numeric address.
And curl is just broken for numeric IPv6 addresses completely:
[dwmw2@i7 activesyncd]$ curl http://%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5D/ curl: (3) [globbing] error: bad range specification after pos 9
I think you get a little further if you put the addresses in /etc/hosts.
David Woodhouse dwmw2@infradead.org writes:
And curl is just broken for numeric IPv6 addresses completely:
[dwmw2@i7 activesyncd]$ curl http://%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5D/ curl: (3) [globbing] error: bad range specification after pos 9
$ curl http://%5C%5C%5C%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5C%5C%5C%5D/ <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
Andreas.
On 06/21/2011 05:54 AM, Andreas Schwab wrote:
David Woodhouse dwmw2@infradead.org writes:
And curl is just broken for numeric IPv6 addresses completely:
[dwmw2@i7 activesyncd]$ curl http://%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5D/ curl: (3) [globbing] error: bad range specification after pos 9
$ curl http://%5C%5C%5C%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5C%5C%5C%5D/
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
Or "curl -g" to disable globbing...
Once upon a time, Josh Stone jistone@redhat.com said:
On 06/21/2011 05:54 AM, Andreas Schwab wrote:
David Woodhouse dwmw2@infradead.org writes:
And curl is just broken for numeric IPv6 addresses completely:
[dwmw2@i7 activesyncd]$ curl http://%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5D/ curl: (3) [globbing] error: bad range specification after pos 9
$ curl http://%5C%5C%5C%5B2001:8b0:10b:1:21d:7dff:fe04:dbe2%5C%5C%5C%5D/
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
Or "curl -g" to disable globbing...
It still can't handle link-local addresses:
$ curl -g 'http://%5Bfe80::230:48ff:fe41:51d5%br0%5D/' curl: (6) Could not resolve host: [fe80::230:48ff:fe41:51d5%br0]; Cannot allocate memory
Lynx and Elinks make the request OK but includes the '%br0' in the Host: header, which lighttpd (at least) answers with "400 Bad Request". Firefox works (it strips the %br0 from the address before putting it in the Host: header).
It would appear that the CLI/terminal web client support for IPv6 link-local addresses is lacking.
On Thu, Jun 02, 2011 at 04:40:10PM +0100, Richard W.M. Jones wrote:
Is there an easy way I can set up IPv6 and a handful of machines on my LAN for testing, without requiring any IPv6 internet connection or an IPv6 assigned prefix?
Yes, every system automatically chooses a link-local address.
The Linux machines on my LAN appear to have acquired IPv6 addresses, eg:
$ ip addr show eth0 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UNKNOWN qlen 1000 link/ether 00:e0:81:74:02:28 brd ff:ff:ff:ff:ff:ff inet 192.168.0.128/24 brd 192.168.0.255 scope global eth0 inet6 fe80::2e0:81ff:fe74:228/64 scope link valid_lft forever preferred_lft forever
fe80:: are link local. The hint is "scope link".
but pinging them gives me strange errors:
$ ping6 fe80::2e0:81ff:fe74:228/64 unknown host $ ping6 fe80::2e0:81ff:fe74:228 connect: Invalid argument $ ping6 fe80::2e0:81ff:fe74 connect: Invalid argument
For using link local addresses, you need to specify the outgoing network interface, because the addresses are global to the host, not local to any one interface. Otherwise, it would have no way of knowing which interface to send the packets out of.
Also, don't put the /nn, because that is the prefix length (equivalent to the IPv4 subnetwork mask).
There are two ways typically to specify outgoing interface for link local:
ping6 -I eth0 fe80::2e0:81ff:fe74:228
or:
ping6 fe80::2e0:81ff:fe74:228%eth0
Some apps work with one vs. the other.
Another tidbit: for IPv6 literals like above, some applications require you to put them in square brackets. For example, web browsers need this syntax:
http://%5Bfe80::2e0:81ff:fe74:228%eth0%5D/
Support for %iface or [] syntax may not be universal, unfortunately.
On Thu, Jun 2, 2011 at 5:40 PM, Richard W.M. Jones rjones@redhat.com wrote:
Is there an easy way I can set up IPv6 and a handful of machines on my LAN for testing, without requiring any IPv6 internet connection or an IPv6 assigned prefix?
If you do want an assigned prefix, or real connectivity and not want to set up a (static) tunnel it may be possible to use 6to4 if your router is not behind NAT. See http://fedoraproject.org/wiki/IPv6Guide#Tunnel_Configuration.
I did not test this myself (in Fedora), and 6to4 tends to be a bit flaky (in general) so ymmv.
Regards, François
On Thu, Jun 02, 2011 at 09:11:47PM +0200, fkooman@tuxed.net wrote:
On Thu, Jun 2, 2011 at 5:40 PM, Richard W.M. Jones rjones@redhat.com wrote:
Is there an easy way I can set up IPv6 and a handful of machines on my LAN for testing, without requiring any IPv6 internet connection or an IPv6 assigned prefix?
If you do want an assigned prefix, or real connectivity and not want to set up a (static) tunnel it may be possible to use 6to4 if your router is not behind NAT. See http://fedoraproject.org/wiki/IPv6Guide#Tunnel_Configuration.
I did not test this myself (in Fedora), and 6to4 tends to be a bit flaky (in general) so ymmv.
I was thinking about the lowest barrier to entry possible so that as many people can test Fedora IPv6 next week, even if they're only testing it on their LAN.
Setting up 6to4 involves at least joining a service like sixxs, which even if free takes a certain amount of time and effort.
I just turned on radvd, and now my Fedora machines are getting IPv6 addresses. By adding DNS entries for a few machines, it's choosing IPv6 protocol by default for many things including writing (if not sending) this email.
Rich.
On 06/02/2011 04:11 PM, Richard W.M. Jones wrote:
Setting up 6to4 involves at least joining a service like sixxs, which even if free takes a certain amount of time and effort.
The method you quoted does not require an account with a tunnel provider.
There is an RFC giving provisions for global IPv4 to IPv6 routers using the 2002 prefix. Sprint has some 6to4 routers spread around the USA.
I'm using the 6to4 method (for several years now). The only requirement is that you must perform the setup on an Internet-connected router box, not on a workstation behind a router.
The other method that you are thinking of hands out native IPv6 addresses.
On Thu, Jun 02, 2011 at 04:53:25PM -0500, Michael Cronenworth wrote:
On 06/02/2011 04:11 PM, Richard W.M. Jones wrote:
Setting up 6to4 involves at least joining a service like sixxs, which even if free takes a certain amount of time and effort.
The method you quoted does not require an account with a tunnel provider.
There is an RFC giving provisions for global IPv4 to IPv6 routers using the 2002 prefix. Sprint has some 6to4 routers spread around the USA.
I'm using the 6to4 method (for several years now). The only requirement is that you must perform the setup on an Internet-connected router box, not on a workstation behind a router.
The other method that you are thinking of hands out native IPv6 addresses.
Given that I mostly don't know about IPv6, what's the best way for people to test IPv6 next Wednesday, given what I think are the following common limitations:
- they'll have one (or two if we're lucky) Fedora machines
- they'll be using a private LAN behind a $40 router that doesn't know anything about IPv6
- they have very limited time and want to do the minimum work possible to set it up
- they themselves know next to nothing about IPv6 ?
Rich.
On 06/02/2011 05:02 PM, Richard W.M. Jones wrote:
Given that I mostly don't know about IPv6, what's the best way for people to test IPv6 next Wednesday, given what I think are the following common limitations:
- they'll have one (or two if we're lucky) Fedora machines
They could connect their Internet modem directly to their Fedora machine.
- they'll be using a private LAN behind a $40 router that doesn't know anything about IPv6
If they're lucky enough to have a DD-WRT (or OpenWRT) supported router, they could flash the firmware to those distributions and use the 6to4 method. DD-WRT has a web-based tool to do the configure work.
- they have very limited time and want to do the minimum work possible to set it up
Unfortunately, if their ISP doesn't provide IPv6, there is not an easy method in Fedora. You either setup a local 6to4 tunnel or create an account with a tunneling provider.
- they themselves know next to nothing about IPv6
The 6to4 method is simple to implement. You just have to have the correct hardware configuration. No specific IPv6 knowledge is required beyond following some directions.
There is another method for tunneling, which Microsoft Windows uses, that involves UPnP (to forward an IPv4 port on the local router). It is called Teredo[1]. Most dumb $40 routers support UPnP by default so Fedora could add Teredo support and provide easy-to-use (default?) IPv6 access.
I think we're talking at cross-purposes here. I want people to be able to test IPv6 *on their local LAN only* next Wednesday with the minimum amount of fuss.
They can just test that it works between two machines, one Fedora, one might be Fedora or it might be something else like Windows acting as a client. This should be sufficient to test a variety of common services: HTTP, NFS, SMB, DNS and that sort of thing.
I don't want to be guiding people through flashing their routers ...
Rich.
On 06/02/2011 05:29 PM, Richard W.M. Jones wrote:
I think we're talking at cross-purposes here. I want people to be able to test IPv6*on their local LAN only* next Wednesday with the minimum amount of fuss.
In that case: Since Fedora defaults to link-local, they would have to run radvd locally just like you had to, run a DHCPv6 server, or assign IPv6 addresses manually in NetworkManager. My suggestion would be to set IPs manually. (fd00:1:2:3::1, fd00:1:2:3::2, etc)
I don't want to be guiding people through flashing their routers ...
Aw, you're no fun. :P
-
Adam Williamson -
Andreas Schwab -
bernd.stramm＠gmail.com -
Björn Persson -
Chris Adams -
Chuck Anderson -
Daniel P. Berrange -
David Woodhouse -
Fabian Deutsch -
François Kooman -
Josh Stone -
Michael Cronenworth -
Richard W.M. Jones