Re: [dnsext] Some thoughts on the updated aliasing draft

[Tony Finch <[e-mail]>]

John Levine <[e-mail]> wrote:
>
> So what can we offer?  If it's something like CLONE or BNAME, we offer
> an upgrade path.  You're no worse off than you'd be with manual
> bundling and manual application configuration, and to the extent you
> upgrade your applications to know about the new DNS stuff, your
> configuration job gets easier.
>
> That's not as cool as the mythical magic hack.  Will people find it
> useful?  Having done my share of SMTP server hackery, I would.  If
> it's as important to handle variant names as people say it is, they'll
> upgrade.  If not, well, that's OK too.

Yes.

I think it's worth splitting the feature into forward and reverse parts.

The forward direction is alias -> canonical, as in CNAME and DNAME. We can
use CNAME synthesis to support old clients, and perhaps a DNSSEC algorithm
bump to avoid validator problems. Old servers can work with the new BNAME
(or whatever) aliases using their existing support for aliases, which
typically means explicit per-alias configuration.

The reverse direction is canonical name -> alias list, which is for the
server to automatically configure its aliases from a trusted part of the
DNS. (Does the client have any use for this information?) Servers that
understand this feature are easier to configure. This feature is also
useful if the aliases no not use BNAME, e.g. if they are CNAMEs or A or
AAAA records.

Both the new forward and reverse features will be useful independently of
each other.

Tony.
-- 
f.anthony.n.finch  <[e-mail]>  http://dotat.at/
Trafalgar: Westerly or southwesterly, becoming cyclonic at times, 4 or 5,
occasionally 6 in Biscay and Fitzroy. Moderate or rough. Rain or showers, fog
patches in north Biscay. Good to poor, occasionally very poor in north Biscay.