IPv6 now live on bitcoin network - please test

[miscreanity]

A static IPv6 subnet with privacy extensions enabled is no worse than a static IPv4 address and NAT. Dynamic IPv4 address don't offer all that much protection of someone really wants to find out who you are; the ISP can always look up who was assigned that address at the time.

For anything more than superficial anonymity, you need onion routing, al la Tor or I2P, to disguise the sources of your connections among a pool of other addresses with multiple independent levels of indirection.

cjdns?

[1713507997]

1713507997

[1713507997]

1713507997

[1713507997]

1713507997

[pc]

Well, you've convinced me to figure out how to compile Bitcoin again. It's gotten a lot easier, so thank you.

I have it running, and it'll make IPV6 out connections (especially if I -addnode them), but I haven't yet received any incoming connections, on either IPv4 or v6. I have UPnP for IPv4, which works in the stock client, so it should be listening okay. Does it advertise both its IPv4 and v6 addresses to the network?

I'm on Mac OS X Snow Leopard 10.6.8, with bitcoin master afdd5941. I have the "privacy" IPv6 mode on, so there's the unchanging MAC-address-generated address, as well as a (daily?) changing random address. The netstat makes it look like it's making the outgoing connection from the random address, but it should probably be announcing the unchanging address, right? Is there a way to tell what it's doing, and what it should be doing?

Also, this is through a 6to4 (2002:) address managed by my router, if that matters.

Thanks.

[Pieter Wuille]

Where is the best place to report compile errors?

http://github.com/bitcoin/bitcoin/issues.

[chufchuf]

Does each IPv6 have a corresponding ipv4 address, or does IPv6 just mean spam will be a lot harder to control?

[LightRider]

Does each IPv6 have a corresponding ipv4 address, or does IPv6 just mean spam will be a lot harder to control?

Ipv6 is a superset of ipv4, meaning that every ipv4 address corresponds to an ipv6 address, but there are far more ipv6 addresses than can ever be represented by ipv4.

Spam is a problem independent of the protocol used to send it. If you want less of it, encourage a society that does not incentivize such profit seeking behavior.

[chufchuf]

Ipv6 is a superset of ipv4, meaning that every ipv4 address corresponds to an ipv6 address, but there are far more ipv6 addresses than can ever be represented by ipv4.

Spam is a problem independent of the protocol used to send it. If you want less of it, encourage a society that does not incentivize such profit seeking behavior.

If IPv6 addresses are a subset of IPv4s, then spammy IPv4 ranges can be blocked without having to look into an insanely large database at every possible spam query.

But I notice by your answer where you say it's not the fault of the technology that this may not be the case. If so then it would be terrible news and allow for tons of spam.

[LightRider]

Ipv6 is a superset of ipv4, meaning that every ipv4 address corresponds to an ipv6 address, but there are far more ipv6 addresses than can ever be represented by ipv4.

Spam is a problem independent of the protocol used to send it. If you want less of it, encourage a society that does not incentivize such profit seeking behavior.

If IPv6 addresses are a subset of IPv4s, then spammy IPv4 ranges can be blocked without having to look into an insanely large database at every possible spam query.

But I notice by your answer where you say it's not the fault of the technology that this may not be the case. If so then it would be terrible news and allow for tons of spam.

Superset, not subset.

[Mike Hearn]

FWIW I have tested this and it worked fine.

[nybble41]

Ipv6 is a superset of ipv4, meaning that every ipv4 address corresponds to an ipv6 address, but there are far more ipv6 addresses than can ever be represented by ipv4.

Spam is a problem independent of the protocol used to send it. If you want less of it, encourage a society that does not incentivize such profit seeking behavior.

If IPv6 addresses are a subset of IPv4s, then spammy IPv4 ranges can be blocked without having to look into an insanely large database at every possible spam query.

But I notice by your answer where you say it's not the fault of the technology that this may not be the case. If so then it would be terrible news and allow for tons of spam.

There are actually several ways to encode an IPv4 address as an IPv6 address, though only one--6to4--is in common use. This version associates a /48 IPv6 subnet with each 32-bit public IPv4 address. To filter these you would extract the IPv4 address and look it up in a normal IPv4 database.

However, most IPv6 addresses will not correspond to any IPv4 address. Only networks which have been assigned an increasingly scarce public IPv4 address can use 6to4 as a substitute for native IPv6, and the remaining IPv6 addresses have no connection to the IPv4 namespace. However, each end user is to be assigned a single, contiguous subnet, ranging from a /48 to a /64. Tracking these subnet prefixes in a database for filtering should be no more difficult than tracking individual public IPv4 addresses, and less prone to false positives due to address sharing (dynamic IPs and NAT). The databases would be somewhat larger, due to the longer addresses and more specific targeting, but not unmanageably so.

[chufchuf]

By somewhat larger, how many more IPv6s can there be than possible IPv4s?

[chufchuf]

Nybble, By 'somewhat larger' what do you mean, how many more IPv6s can there be than possible IPv4s?

[Pieter Wuille]

By somewhat larger, how many more IPv6s can there be than possible IPv4s?

79228162514264337593543950336 times more.

[nybble41]

Nybble, By 'somewhat larger' what do you mean, how many more IPv6s can there be than possible IPv4s?

It's not a matter of how many more IPv6 addresses there are in absolute terms. The database would only store prefixes, so the increase in table entries is just from having individual prefixes for each end-user where currently you might block an entire ISP due to dynamic IPs or high-level NAT. This is the price for avoiding the false positives which are inevitable with IPv4, but not with IPv6. If you already have the end user's IPv4 address in the table, or don't care about the existing false positives, then there is no change in the number of entries.

However, IPv6 addresses are 128-bit rather than 32-bit as in IPv4, so each table entry would go from five bytes to 17 bytes, including one byte in each case for the prefix length. That's assuming you use a flat table with fixed record sizes rather than prefix compression or similar.

NOTE: You could cut the IPv6 record size down to 9 bytes by assuming the prefix is no longer than 64 bits, which is reasonable since each end-user is supposed to have at least a /64, if not a /56 or /48.

[Serenata]

Glad to hear that Bitcoin has gone IPv6! Unfortunately I don't have IPv6 equipment just yet, to test but soon (I hope...)

[mb300sd]

How does bitcoind select which ipv6 address to advertise and listen on for incoming connections? I have 3-6 public ipv6 addrs on each computer, but only the one assigned by my dhcpv6 server (via a reservation) has ports open on the firewall.

[Pieter Wuille]

How does bitcoind select which ipv6 address to advertise and listen on for incoming connections? I have 3-6 public ipv6 addrs on each computer, but only the one assigned by my dhcpv6 server (via a reservation) has ports open on the firewall.

You may want to look at the new -externalip option. It tries to be smart by default, but if you have several addresses in the same network, it can fail.

[mb300sd]

How does bitcoind select which ipv6 address to advertise and listen on for incoming connections? I have 3-6 public ipv6 addrs on each computer, but only the one assigned by my dhcpv6 server (via a reservation) has ports open on the firewall.

You may want to look at the new -externalip option. It tries to be smart by default, but if you have several addresses in the same network, it can fail.

The problem is that I have a dynamic IPv4 and static IPv6. It seems like I don't get any incoming IPv4 connections if I set externalip to an my v6.

Fixed by disabling use_tempaddr on the interface, now it only has 1 ipv6.

[TeslaUa]

I already got IPv6 working here but I'll wait until this feature will gets into official release. Compiling just isnt my thing but I support the move to IPv6

[TehZomB]

Fixed by disabling use_tempaddr on the interface, now it only has 1 ipv6.

Can you explain how you did that?

Thanks

[mb300sd]

Fixed by disabling use_tempaddr on the interface, now it only has 1 ipv6.

Can you explain how you did that?

Thanks

sysctl net.ipv6.conf.eth0.use_tempaddr=0