[Freedombox-discuss] Distributed naming, again

Clint Adams clint at debian.org
Mon May 16 15:57:17 UTC 2011


I am going to anger Jonas by discussing something that
interests me, and even though UX is the very most
important thing about the FreedomBox project, it is
only a small part of the text below.

A group of us got together recently to discuss the state
of distributed naming technology.  We talked about
Zooko's Triangle[0], Netsukuku[1] ANDNA, i2p[2],
PetNames[3], Namecoin[4], and Tom Lord's proposal[5a,5b].

We omitted dot-p2p, IDONS, Tor Hidden Services.

Some questions were raised, which hopefully will be
communicated to the appropriate mailing lists shortly.

The part I would like to get into the mailing list
archive before it is lost is this ASCII art diagram
expressing what some people think is necessary from
a system or a pair of systems working in tandem:

          nickname    (legacy) global
               \         /
[secure backup] \       / [control retention]
[introductions]  \     /
[intelligible UX] \   /   [resource conservation]
                   \ /
                   key
                    |
                    | [reachability over arbitrary transport]
                    |
                  address

Starting from the lower half, there should be a mechanism
to look up a public key and get one or more addresses.
The address specification should be flexible; in other words,
the destination should not be restricted to a member of
a particular overlay network.  The public key should enable
the cryptographic verification of the destination/peer.

There should be two different methods to map a name to a key:
one where names are meaningful only locally, and one that can
be used through a DNS gateway.

For the former, there should be a methods of secure backup,
so the name mappings are not lost when a person's trusted
computer experiences data loss; introductions, so we can
find the keys we are looking for; and intelligible UX, so
we can record the mappings in a sane and pleasant way.

When names are locally meaningful there is no need for
reservation and renewal, but in the other case, there
needs to be a process for ownership and maintenance
thereof.  This should not require extensive resource
usage; if my small, low-power computer needs to burn
enormous amounts of CPU, storage, or network traffic,
it will not be able to retain its name.

[0] http://zooko.com/distnames.html
[1] http://netsukuku.freaknet.org/
[2] http://www.i2p2.de/
[3] http://www.skyhunter.com/marcs/petnames/IntroPetNames.html
[4] http://privwiki.dreamhosters.com/wiki/Bitcoin_DNS_System_Proposal
[5a] http://lists.alioth.debian.org/pipermail/freedombox-discuss/2011-April/001474.html
[5b] http://lists.alioth.debian.org/pipermail/freedombox-discuss/2011-April/001475.html



More information about the Freedombox-discuss mailing list