To:
itojun@iijlab.net
Cc:
Bill Manning <bmanning@ISI.EDU>, dnsop@cafax.se
From:
Erik Nordmark <Erik.Nordmark@eng.sun.com>
Date:
Wed, 19 Sep 2001 14:02:10 +0200 (CEST)
In-Reply-To:
"Your message with ID" <12633.1000889185@itojun.org>
Reply-To:
Erik Nordmark <Erik.Nordmark@eng.sun.com>
Sender:
owner-dnsop@cafax.se
Subject:
Re: operationally (if not yet WG) related
> > >>"We recommend DNS administrators DO NOT use mapped IPv6 addresses with > >>opcode 28 <type AAAA> resource records in ANY zone file." > > > > I agree 120%. I also agree with Bill's suggestion for DNS operation. > more words: > from what RFC2553 3.7 says, IPv4 mapped addresses are not supposed to > appear on the wire. they are OS-internal representation of IPv4 peers > on top of AF_INET6 socket. they are not supposed to appear in the wild > like AAAA records. I think it would make sense to figure out whether it makes sense to extend the restriction on the IPv4-mapped address format to other places where it would appear in essentially textual form. While IPv4-mapped addresses are quite useful in the API I have not seen any need to have them appear in textual format e.g. in configuration files. Thus instead of having configuration files (or other textual literal addresses) use ::ffff:a.b.c.d they should just use a.b.c.d. It makes sense discussing the general (non-DNS specific) issue in ipng. For example, I can envision getting implementors modify getnameinfo and inet_ntop to always present an IPv4-mapped address as the IPv4 address i.e. without the leading "::ffff:" text. That would help prevent these addresses appearing in undesired places in the future. (Of course, one could also entertain having getaddrinfo/inet_pton refuse to *parse* such addresses but that might be more difficult to introduce immediately.) Erik