A DNS question re 6to6/IPv6 host IN A records.

[itojun@iijlab.net]

	I don't think freebsd-stable is suitable for this thread, so dropped
	it from cc: list.

>I don't agree.  There may be differences what one means with 'production' 
>though.  Personally, if I had a power to switch on IPv6 on www.google.com 
>hosts, I would only do it by adding www.ipv6.google.com: NOT with 
>www.google.com.  People who are afraid of degrading service and it costing 
>real money are reluctant.

	it depends on how you run your IPv4/v6 servers.  for instance, we are
	running ftp.iij.ad.jp, one of the most famous anonymous ftp server in
	Japan, dual-stacked.  this is because we think it robust enough.

	for our company website, www.iij.ad.jp, we do like this:
	the we are using is like this:
	- "www.iij.ad.jp" has A record and AAAA record.  there actually are
	  two machines.
	- "A" record for www.iij.ad.jp points to one of them, which runs very
	  stable version of IPv4.  the machine has the actual data.
	- "AAAA" record for www.iij.ad.jp points to another one, which NFS-
	  mounts the data partition from IPv4 one.
	we do monitor them closely, and they have impressive uptime.

	if you run www.ipv6.google.com, people won't be able to smoothly
	migrate to IPv6.  i believe it a major drawback.

	my point is, there are a lot of ways you can operate dual-stack
	services, and you just need to make a right choice.

>An example of potentially technically capable stuff: IPv6 service of 
>playground.sun.com/ipv6/ was down/out-of-sync some time ago.. and that 
>isn't even a "production" site.

	i guess this is because IPv6 variant of playground.sun.com has not
	been given enough babysitting/monitoring.  I don't like it to be seen
	as an example of lousiness of IPv6.  it is lousy operation that causes
	trouble, not lousy protocol/implementation.
	(i'm not blaming those who administer playground.sun.com)

itojun