Nicolai writes:
> I'm wondering what the roadmap/plan is for DNSCurve.
Step 1: Convince people that metadata is private information and needs
to be encrypted by default. Oh, maybe this step is handled now. :-)
The three main deployment targets for DNSCurve are
* caches, preferably on user laptops/smartphones rather than ISPs;
* leaf servers; and
* higher-level servers---roots are best run on user machines, but
protecting .com is a more interesting challenge.
Obviously all of these can benefit from broader software support, and
own emphasis at the moment is on simplifying this by improving the ease
of use of the underlying crypto tools.
The DNSCurve protocol per se is stable, but there's a lot to be said
about improving DNS: for example, I don't see why someone looking up
"
rites.uic.edu" should send anything more than "edu" to the root name
servers. On the operational side, I'd like to see more administrative
tools with a clear separation between
* local configuration on a single administrator machine (whether
through command-line tools or a web interface) and
* replicating the configuration to all of the administrator's servers
(for example, with unison)
using the filesystem as the underlying database. For DNSCurve this would
normally mean generating a single key and replicating that key to all of
the administrator's DNS servers---preferably with a single name using
multiple IP addresses, although multiple names would also be useful for
compatibility with parents that require a single IP address per name.
> I also note that DJB hasn't enabled it
> on his recursive servers for
cr.yp.to, which is surprising and strange.
Some of my caches use it, some don't; I manage something like 100
machines in my spare time (and a similar number of caches, with many
different services), so upgrades often take me a while, and I have a
strong preference for tools that are _really_ easy to use instead of
just close.
---Dan