A small response and some broader context for those interested.

On 02/19/2014 08:16 PM, Benjamin Cordes wrote:
> some good points here.
>
> * on the international level we have irreversible SWIFT, yes. but,
> SWIFT is owned by an international regime. so its very unlike the
> internet as a network. its an important tool for sanctions etc. you
> can't just join the network. if you're very rich you might get a
> license and join as a member. up to now the international financial
> network and the internet evolved separately.
>

Yes, I wasn't trying to say it is easy to get irreversible payments
(were you responding to my message? I'm not sure, you didn't quote it),
I was trying to illustrate the way the system is layered with gradations
of finality.

Perhaps my message was over long and I should have focused on what I
consider to be the key point: networks like SWIFT are intended to allow
certain actions, but not others, but the nature of information transfer
is that its intent is not explicit. That's why hacking, in the purest
sense, is possible. Consider the very earliest forms of hacking
("phreaking" etc.) - the telephone network was a controlled information
transfer network, just like SWIFT is, with fixed purposes. Eventually
people found ways to use that network for purposes other than those it
was designed for.

Data is not meaning without context, it is just data. If we remove the
context, we remove the meaning. Thus, a SWIFT transfer between Alice and
Bob might be meaningless, while two SWIFTs between Alice and Bitstamp
and Bitstamp and Bob have context and have meaning.

This line of thinking leads to the idea of adding layers of abstraction
to transfer. My first concept in this area was along those lines - make
transfers P2P and even over multiple hops to obscure their meaning. I
wrote the idea in this message:
http://www.reddit.com/r/Bitcoin/comments/1doeih/idea_for_solving_the_exchange_problem_just_add/

... before realising the difficulty of proving wire transfers, and have
been working with dansmith for most of a year on how to achieve this in
a complete, working system. We even came to Milan in December to discuss
it but it didn't seem many people were interested (nice to see that's
changing!)

Our latest ideas have diverged slightly, but both versions can be found
on the last page of the ssllog thread starting around here:

https://bitcointalk.org/index.php?topic=173220.msg3903951#msg3903951

Obviously reading the whole thread gives a lot more context.

and my current codebase is here:
https://github.com/AdamISZ/ssllog/tree/fullthreeparty (still a lot of
gunk in there, but it might give people some ideas).