[12622] in Commercialization & Privatization of the Internet
Re: About the creation of new 3-letter domain(s)
daemon@ATHENA.MIT.EDU (Barry Shein)
Tue May 24 18:59:26 1994
Date: Tue, 24 May 1994 12:46:11 -0400
From: bzs@world.std.com (Barry Shein)
To: barney@databus.com
Cc: com-priv@psi.com
In-Reply-To: <9405232251.AA11821@databus.databus.com> (message from Barney Wolff on Mon, 23 May 94 22:51 EDT)
>From: Barney Wolff <barney@databus.com>
>Barry, I'm confused about which pieces of software would have to be aware
>of the properties of your suggested phone-based domains:
> any program which uses domain names
> my local resolver
> my local nameserver
> all root nameservers
> the nameservers for the new phone-based domains.
>
>Tell me it's just the last one, please!
There isn't a lot to be aware of, just which order the number was
typed. Probably the best place would be in the common software library
that the user interfaces use, and any part of the resolver or domain
system itself will only see "DNS normal order" (reverse of how we
normally type phone numbers to coincide with the general hierarchical
ordering DNS uses.)
But, using the principle of "be conservative in what you send and
liberal in what you accept" there'd certainly be no HARM in other
pieces of software knowing about this, it's just that it would neither
be necessary nor promised. Sort of like we have things set up here (as
do many sites) so you can type hostname.bitnet and the right thing
happens in the MTA, even if it would be better if no one ever typed
anything like that, in some sense.
But it would only be required in the user interface software or
middleware perhaps (common subroutine library.) Actually it's not
required at all, it's just a human convenience, you can type the phone
number domains in "DNS normal order" if your system has absolutely no
knowledge about this (or alternatively use your mailer program's
address book to do this like one would now.)
In most modern Unix'es, as an example, the sensible place to put the
software might be the resolver library, routines like gethostbyname()
already have a field in their return value for the "official" name of
the host which could always be the DNS normal order presentation.
This is completely analogous to the current use of the environment
variable LOCALDOMAIN (see resolver(3) on a Unix system that has the
resolver routines.)
>But I would have thought that my local nameserver is likely to cache
>stuff from 002.1234.5678.fon and so go looking in the wrong place for
>987.6543.5678.fon. And it is likely to be confused if a nameserver for
>.fon tells it to go look in 002.fon for 002.1234.5678.fon. Ok, tell me
>where I'm going wrong, again :-).
Well, in the first place, using a .fon (or similar) is extraneous.
That's something I am trying to avoid so we can keep it to the
shortest possible string (i.e. a domain of all digits is to be
interpreted as a phone number domain, numeric IP addresses shall
henceforth be typed to UA software with square brackets, as currently
required by RFC822 but extended beyond just email addresses.)
I don't understand your example, what's the relationship between
002.1234.5678 and 987.6543.5678? I don't think there's any confusion
caused by caching, there's really no difference between this scheme
and the one we currently use, other than we're trying to make some
provision for entering domains in either order unambiguously.
-Barry Shein
Software Tool & Die | bzs@world.std.com | uunet!world!bzs
Purveyors to the Trade | Voice: 617-739-0202 | Login: 617-739-WRLD
