[messaging] GNU Name System

[George Kadianakis]

Mike Hearn <mike at plan99.net> writes:

>>
>> GNS has really nice features, but like any other cryptographic system I
>> think
>> the hard part is how to make it easy to use.
>>

Indeed.

GNS is a fun project :)

Looking forward to see how GNUNet will use it.

> Yes, GNS seems to hit problems at this point in the paper:
>
>    "Bob gets to know Alice in real life and obtains her public key"
>
>
>> The web usage of most non-nerds around me is that if they want to go to
>> facebook
>> for example, even though they visit it every day several times, they type
>> in the
>> search engine (usually google) 'facebook' and follow what the search engine
>> dictates what is facebook.
>
>
> Funny fact - one of the top searches on Google is "yahoo" and one of the
> top searches on Yahoo is "google". Or at least it used to be. People
> navigate to search engines using search engines too. And why not? A search
> engine is miles better than a URL bar for ordinary users. It does spelling
> correction, understands non-English alphabets, and doesn't force you to
> think about the distinction between an address and what you actually want.
> Most importantly, it *always* works and never gives you mysterious errors.
>

Good points :)

> W.R.T. the utility of censorship free naming, I'm not sure a naming only
> system is actually that useful. DNS is already decentralised across all
> countries. Sites that have hit DNS censorship in the past have basically
> always been successful at playing jurisdictional arbitrage. For something
> like GNS to be useful you'd need a web site that can't get a domain name in
> any country or TLD, presumably due to illegality (what else can cause
> this?), yet doesn't mind exposing its IP address in the clear.
>
> In practice, sites that face such across the board levels of censorship
> i.e. Silk Road and friends all need to hide their server location as well,
> in which case they end up just using Tor for everything including naming.

FWIW, Tor has also been planning to use a similar key blinding scheme
for HS names. This is done so that HSDirs can't harvest the names of
published or requested HSes. You can read more about it here:
https://gitweb.torproject.org/torspec.git/blob/HEAD:/proposals/224-rend-spec-ng.txt#l1635
https://trac.torproject.org/projects/tor/ticket/8106
And for those of you who enjoy provable security, you can find a
sequence-of-games proof here:
https://lists.torproject.org/pipermail/tor-dev/2013-December/005943.html

Of course, from a quick skim of the paper, GNS is much more than a
query privacy scheme. For example, I wonder if Tor could use the
petname system of GNS to help make HS names human-memorable.
.oO(But does that mean that another DHT needs to be added to the
protocol? Need to read the paper properly...)