On Mon, 29 Feb 2016, Dave Crocker wrote:
Normally we try to leave "private use" ranges in registries for people
to experiment on. It would be good to have that here as well, or else we
won't be able to differentiate experimentation from standarization. I
suggest reserving the double underscore space (__*) for private use
(provided this isn't already in wide use)
There is a real problem with private naming, and that is that any use of
private names that becomes successful is then faced with having to change all
existing behaviors to move to a publicly-registered name.
It is still much preferred. If you have two competing proposals both
using _foo in experimentation, you're in even more trouble. Most regular
registries work with private use. Yes, there are examples of some that
are now tainted forever, but I do believe it prudent to ensure that we
can point to some text stating the implementors made the wrong choice.
The document should probably explain the RRtype listed in the registry
and interaction/allowance with CNAME/DNAME.
1. What do you mean 'explain' and why should the registry contain this
information? Specifically, how is that information essential to simple and
useful operation of the registry?
The RRTYPE listed do not include CNAME. So are CNAME's disallowed? That is
the question I want to be clear for implementors. I am not saying to
dictate whether or not those are allowed, that is indeed for the
individual documents to say. I just want some generic advise to
implementors so they don't blindly reject CNAME/DNAME based on the list
of RRTYPE's in this registry.
What are the requirements for entry into this registry. I would not want
to see a rush of people registering vanity names for pet projects,
taking away all the sensible one word entries. I see _mail is available :)
Well, we know the concern about vanity use of DNS-related names is a long way
from silly, so alas I guess we have to worry about that. grrr...
From the list of IANA choices for this:
http://tools.ietf.org/html/rfc5226#section-4.2
I'm inclined to suggest 'Specification Required'. In my own view, an
internet draft ought to qualify, since they no longer disappear, but I
believe the community view is that it would require an RFC or the like.
I'd say so too.
Note that as others noted, this listing could also be added to an
existing registry. I haven't thought deeply about which of the two
options should be prefered, but a discussion of this draft will
hopefully make that clear.
Paul
_______________________________________________
DNSOP mailing list
DNSOP@ietf.org
https://www.ietf.org/mailman/listinfo/dnsop
