Re: [DNSOP] why classes are useless, was New Version Notification for draft-sullivan-dns-class-useless-01.txt

[Paul Vixie <[e-mail]>]

Andrew Sullivan wrote:
> On Fri, Mar 18, 2016 at 06:44:44PM -0400, Michael StJohns wrote:
>> responses within the UDP sizes.    The class field might have been a useful
>> way to do that, especially for things related to keys and signatures.
>
> There are lots of things the class field _might_ have been useful for.
> What I've been trying to suss out in this conversation is whether
> there is any way to make that potential utility actual.  Do you think
> the repurposing you're mentioning is a realistic hope, or just a road
> not taken?

my own view is that it's impossible to negotiate new data patterns end 
to end. EDNS allows negotiation hop by hop, but recall that extended 
label types was removed because it would have allowed data to enter the 
system in authorities which could not be expressed other than hop by hop 
(to the first recursive in the data path.)

thus, we used the class field in the OPT RR for hop by hop signaling, 
but there is no analogous end-to-end use whereby a zone or an rrset 
might be extended to have another dimension, that of class.

-- 
P Vixie