In message
<CACweHNCAi7JcOW9CX=6FViv1wUoe5fhn7deJ2eieP2-D_FhaSA(_at_)mail(_dot_)gmail(_dot_)com>,
Matthew Kerwin writes:
On 5 July 2017 at 10:02, Mark Andrews <marka(_at_)isc(_dot_)org> wrote:
Who owns a name is a different question to what machines serve the
<name,type,class> tuple and how do you reach those machines. There
is absolutely no reason why the zones <name,IN> and <name,CLASS56>
need to be served by the same machines. There is a argument for
them both being under control of the same people.
Mark
Hi, I'm jumping in at a random time with a possibly dumb question, but
the talk of <name,type> and <name,type,class> tuples got me wondering
about representation in general, and URLs in particular.
RFCs 3986 and 7230 say[*] that every 'host' in a HTTP URL that looks
like a DNS name is a DNS name, and that they have to be resolved to IP
addresses if you want to fetch them, but they don't talk meaningfully
about how to do that resolution. Given that we always assume class=IN
(not to mention type=A|AAAA via happy eyeballs), how would we go about
practically presenting an alternative class in things like URLs?
(Registering a new "alt-http" URL scheme doesn't strike me as a great
idea.)
mailto: is tied to <MX,IN> then <A,IN> and <AAAA,IN> directly or indirectly.
http: is tied to <A,IN> or <AAAA,IN> and perhaps in the future <SRV,IN>
Note the linkage is not in the name but in the definition of the
scheme.
If some scheme needs <type56000,CLASS100> it will be defined in
that scheme and you will call the resolver with <name,type56000,CLASS100>
to lookup data. That may then result in a call to lookup
<name2,AAAAA,IN> to get the addresses of the servers based on the
data returned.
Remember we call resolvers most of the time with <name,type,IN>
today. Changing IN to something else is not hard. You just need
to know to do that and when people write the code to support the
scheme they will do that.
When a new globally resolvable class becomes active whois will show
a set of nameservers for class IN and a set of nameservers for class
FOO. They may be the same or they may be different. One set may
be empty.
Remember this in not new stuff. HS was used this way but without
central delegations. You were still expected to use the namespace
delegated to you. The recursive servers knew how to locate the HS
data. getpwnam() knew how to map from user name to <domain name,
TXT, HS> and lookup the data by calling the resolver with those
values.
Because it's all well and good setting up your own .org hierarchy
under class=FOO or whatever, but there's not much point if you can't
send people to www.not-icann.org using it. Unless you don't want to
expose your new hierarchy to the web ...?
Cheers
[*] https://tools.ietf.org/html/rfc3986#section-3.2.2 :
"""A registered name intended for lookup in the DNS uses the syntax
defined in Section 3.5 of [RFC1034] and Section 2.1 of [RFC1123]."""
I read that as: "if it matches RFC1034 (and isn't overridden by the
specific URI scheme's rules) it's a DNS name." It could be read the
other way, but that just adds more assumptions.
--
Matthew Kerwin
http://matthew.kerwin.net.au/
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: marka(_at_)isc(_dot_)org