"adding IPv6 glue to root zone" document
At RIPE 46 we presented measurements about the effects of adding IPv6 glue to the root zone. As promised, we made more measurements and have written a document: http://www.nlnetlabs.nl/ipv6/publications/v6rootglue.pdf rvdp
Nice! I have two follow-up questions: (1) Some DNS servers have added AAAA for their servers. If they run slave servers for some TLD's, those TLD's "suddenly" have AAAA records for one of their nameservers. Maybe these cases are the real world "experiments" we should look closely at? Have we heard any complaints? I am especially thinking of ns.ripe.net which have one A and one AAAA record. (2) In Sweden we are discussing adding AAAA to the zone, and have also made some studies (not run by me, I am a passive participant in the practical tests). We have tested things like possible problems resolver libraries have to send queries to a nameserver IF the nameserver have both A and AAAA records, the resolver have Ipv6 enabled, but not IPv6 connectivity to the nameserver. Will the resolver fall back to IPv4 transport? It should, and what we have found so far is that it does. I hope we can have some data before the next RIPE meeting. Has anyone else looked at potential problems like this? paf On 2003-10-20, at 15.28, Ronald van der Pol wrote:
At RIPE 46 we presented measurements about the effects of adding IPv6 glue to the root zone.
As promised, we made more measurements and have written a document: http://www.nlnetlabs.nl/ipv6/publications/v6rootglue.pdf
rvdp
Hi, Thanks very much for that document. It is very useful for the community. I have only 2 comments concerning Appendix A: 1) Among the 39 mentioned name servers, why are there duplicates (same names with the same addresses, see for example d.dns.jp (3 times) or ns.nic.ir (2 times) ? 2) As far as I know, the IPv4-mapped IPv6 addresses are irrelevant in this context (see *.dotdm.net et *.netdns.com), do these name servers have regular (2001 or 3ffe) IPv6 addresses ? On the other hand, I would like to share with you the follwing document which focuses on DNS response size calculation (from root servers) and name compression for TLDs with IPv6 taken into account. A part of the document sections deal with the general case, so they can apply to any TLD. The remaining sections contain calculations for .FR and may be easily extended to any other TLD. If you have any comment, suggestion or correction, don't hesitate to send them to: ipv6tech@nic.fr. Please find the document at: http://w6.nic.fr/dnsv6/dns-resp-size-and-name-compression Regards, Mohsen. On 20 Oct, Ronald van der Pol wrote: | At RIPE 46 we presented measurements about the effects of adding | IPv6 glue to the root zone. | | As promised, we made more measurements and have written a document: | http://www.nlnetlabs.nl/ipv6/publications/v6rootglue.pdf | | rvdp
On Thu, Nov 13, 2003 at 17:06:14 +0100, Mohsen Souissi wrote:
Hi,
Thanks very much for that document. It is very useful for the community.
I have only 2 comments concerning Appendix A:
Thanks for your comments.
1) Among the 39 mentioned name servers, why are there duplicates (same names with the same addresses, see for example d.dns.jp (3 times) or ns.nic.ir (2 times) ?
Oops. Seems like I forgot a uniq(1). The result with respect to drops still stands, but the 39 figure is actually less.
2) As far as I know, the IPv4-mapped IPv6 addresses are irrelevant in this context (see *.dotdm.net et *.netdns.com), do these name servers have regular (2001 or 3ffe) IPv6 addresses ?
We queried the auth nameservers for those. They only report mapped addresses, no global addresses. Why are these mapped addresses irrelevant in this context? They show up in the additional section.
On the other hand, I would like to share with you the follwing document which focuses on DNS response size calculation (from root servers) and name compression for TLDs with IPv6 taken into account. A part of the document sections deal with the general case, so they can apply to any TLD. The remaining sections contain calculations for .FR and may be easily extended to any other TLD. If you have any comment, suggestion or correction, don't hesitate to send them to: ipv6tech@nic.fr.
Please find the document at:
I will have a look at it. rvdp
On 14 Nov, Ronald van der Pol wrote: | On Thu, Nov 13, 2003 at 17:06:14 +0100, Mohsen Souissi wrote: | | > Hi, | > | > Thanks very much for that document. It is very useful for the | > community. | > | > I have only 2 comments concerning Appendix A: | | Thanks for your comments. | | > 1) Among the 39 mentioned name servers, why are there duplicates (same | > names with the same addresses, see for example d.dns.jp (3 times) or | > ns.nic.ir (2 times) ? | | Oops. Seems like I forgot a uniq(1). The result with respect to drops | still stands, but the 39 figure is actually less. | | > 2) As far as I know, the IPv4-mapped IPv6 addresses are irrelevant in | > this context (see *.dotdm.net et *.netdns.com), do these name servers | > have regular (2001 or 3ffe) IPv6 addresses ? | | We queried the auth nameservers for those. They only report mapped | addresses, no global addresses. Why are these mapped addresses irrelevant | in this context? They show up in the additional section. ==> Hmmm... Can you reach them? ;-) I can't ! Going back to a discussion which took place 2 years ago on dnsop mailing-list, I found the following thread: http://www.cafax.se/dnsop/maillist/2001-09/msg00021.html Hope that helps... Mohsen.
On Fri, Nov 14, 2003 at 18:30:27 +0100, Mohsen Souissi wrote:
==> Hmmm... Can you reach them? ;-) I can't ! Going back to a discussion which took place 2 years ago on dnsop mailing-list, I found the following thread:
I guess we are not on the same frequency yet. All I am saying is that those mapped addresses are in DNS now. So, they are inserted into the additional section and may cause drops. I agree with you that it is quite useless (and even harmful) to put them in DNS. Are we in sync again? rvdp
On 16 Nov, Ronald van der Pol wrote: | On Fri, Nov 14, 2003 at 18:30:27 +0100, Mohsen Souissi wrote: | | > ==> Hmmm... Can you reach them? ;-) I can't ! Going back to a | > discussion which took place 2 years ago on dnsop mailing-list, I found | > the following thread: | > | > http://www.cafax.se/dnsop/maillist/2001-09/msg00021.html | | I guess we are not on the same frequency yet. All I am saying is | that those mapped addresses are in DNS now. So, they are inserted into | the additional section and may cause drops. I agree with you that it | is quite useless (and even harmful) to put them in DNS. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ==> That's why I was talking about relevance of these addresses in my first message :-) On my turn, I agree with you can put any IPv6 address in the DNS provided that it is correctly formatted. | Are we in sync again? ==> Yes we are! Mohsen.
participants (3)
-
Mohsen Souissi -
Patrik Fältström -
Ronald van der Pol