v4 given v6

[Peter Curran]

Bob


>Might it be a good time to try to go away from this convention (i.e., a
>separate ipv6 subdomain)?
>
>As we try more and more to act like a real dual v4/v6 environment it would
>be best to not do this.
>
>Can advocates of maintaining this convention please state their reasons for
>this to the list?
>


The original reason for adopting this convention (as I think is mentioned in
an item on the web site) is that it gets around the problem of older DNS
implementations that may not have support for AAAA.

In practice I think that there are a couple of good reasons to maintain the
convention.  One has been given by Bertran Buclin (i.e. not mixing
'experimental' and 'production' environments).

My reason is equally practical:

I have a number of v6 systems.  Not all of them are dual-stacked.  On some
of the dual-stacked systems, not all of the services have been ported to v6.

If I have a dual-stack client then I need some way to discriminate between a
v4-only app, a v6-only app and a v4/v6 app.  The mapIPv6 resolver option
doesn't help me here - it is an all or nothing solution.  If I want to
access a service on a v4/v6 node and I know the service is v4-only than I
can specify the 'regular' name and as an A record is retreived then I can
successfully access the service.  Conversely, if I know the service is
v6-only or v4/v6 then I specifiy the ipv6 subdomain and I get an AAAA
record.

So, I think that the concept of a parallel IPv6 domain has some continuing
value for me until all of my applications are bilingual.

My 2p

Peter Curran
TICL/UK