Re: DNS Blockchain Use Case

On Wed, Mar 15, 2017 at 6:47 AM, Melvin Carvalho <melvincarvalho@gmail.com>
wrote:

> Namecoin is a lookup mechaism via a block chain, which was a clever idea.
> But it's not a particularly secure block chain, it's merge mined and there
> has known attack vectors.
>

Yes, systems based on Proof-of-Work are only secure in as much as they have
not been catastrophically attacked by nationstate/APT-level threats.
Furthermore, not only are not only insecure, but fail to live up to modern
best practices of distributed systems design, and fail to achieve the goal
of Byzantine agreement.[1]

Bitcoin and its ilk fail to tolerate network partitions. They will happily
accept writes during a partition then clobber them when the partition heals.

Security-wise this manifests in MitM and eclipse attacks. An attacker with
control of the network can MitM registration, preventing precommitments
from reaching the Namecoin P2P, then stealing names when the precommitment
is revealed to a malicious fork. I would say it's curious this presentation
omits the "secure" leg of Zooko's triangle, but given Namecoin's many
(known) security problems, I can't say I'm *that* surprised they would
downplay security.

But beyond that, this presentation speaks to an ideology which is either
ignorant of or ignores the actual operational aspects of a production
naming system, "namely" (pardon my pun) disputes. The presentation
repeatedly decries "nondeterministic humans" ignoring that dispute
resolution is a nondeterministic process. In the Namecoin world, dispute
resolution is simply unsolved: anyone who perpetrates an account takeover
owns your name thereafter, and there's nothing you can do. This is a total
and complete nonstarter for actually productionizing such a system.

I think there might be value in using decentralized, BFT databases among
the various registrars who share the registries for particular TLDs, but
that's merely an implementation detail. There is absolutely no reason any
of that complexity needs to be exposed to DNS clients or caches, nor does a
system based on a Bitcoin-like PoW scheme even remotely begin to make sense
as a replacement, from many perspectives including fitness for the purpose,
performance, scalability, and security.

[1]: https://eprint.iacr.org/2014/765.pdf