On 03/17/2014 03:24 PM, Manfred Karrer wrote:
>
> To go in that direction the arbitration will become more and more central and powerful as we move all those trust problems over to the arbitration system.
> Its true that with MultiSig there are no theft risks (beside collusion),
I should emphasise for those reading this thread (I know you know
Manfred): the main defence against collusion is that using a randomising
algorithm, a buyer or seller does not know in advance which member of
the pool of arbitrators will handle their case.

but other new problems might to be problematic and hard to solve:
> - Entry costs (to become an arbitrator) need to be low so it will not lead to monopolies

I somewhat disagree; the arbitrator doesn't incur costs, rather
collateral (so, technically, opportunity cost). I think it is necessary
that the collateral is at least of the same approximate size as a
"large" transaction for the system, since otherwise there is incentive
to collude.

> - The max. arbitration volume might be pegged to the collateral, so that might lead to centralisation (a rich arbitrtor can take more jobs as he paid more collateral in, and get richer with doing more jobs)

The collateral should be fixed for a single arbitrator pool, not
variable according to pool members' wishes. This is crucial - it's
essential that each participant in the pool has the same status as all
others, as they will be accessed at random by users.

> - Legal status of arbitrators can turn out that they have to be registered businesses with all the implications (regulation, governmental control,...)

So, we must distinguish two points: since the arbitrator never receives
fiat funds from the counterparties (or anyone else), there should not be
a legal issue. If we conclude that arbitrators still take a legal risk
due to receiving transaction fees in BTC, then that problem has to be
dealt with much as it is currently in the BTC world - different
geographies, different legal systems, partial or full anonymity (as I
said in the doc, the identity used to set up the arbitrator could just
as easily be a PGP identity as a government issued one).

> - Prevent cartel building and influence of bigger companies having many pseudo-independent sub-contractors as arbitrators (see real life examples like the real estate agency Remax,…)
> - Fee policy could lead to kick out competitors (monopoly building)

These two points, and most of the previous ones are basically the same
point - that this kind of system leads to some quasi-centralization. I
think this is at least partially true. And I don't think you will be
able to build an arbitrator-based system without this problem (because
the system still requires trust, however slight). This is mitigated only
by openness and organic growth. Any user can become a member of the
arbitrator pool, as long as they (a) post the collateral, (b) post an
appropriate identity publically and (c) get agreement from the other
members.

But just as critical is that there should be multiple pools. If a pool
starts to get a bad reputation, users are free to choose a different one
(or because of different fees).

> - Sock puppets cannot be prevented completely. Even with real life ID verification a company can hire many arbitrators using their ID but in fact the company has full control about all those arbitrators.
The possibility of Sybil at arbitrator level is one of the main reasons
that the pool needs to be seeded with "real" identities - but here real
does not mean, say, government identities. It means identities with
pre-existing community trust (PGP WoT and other sources). But it need
not be so many - perhaps 3-5 such identities would be fine to start with.
The collateral also dissuades Sybil-ing the pool, but in itself it
definitely is not enough.

> - More exposed to DDos attacks as the arbitrators are limited and well known nodes
Absolutely; back to the centralization point. Any not pure P2P system
has this problem. However, a system of pooling has the advantage that,
as it gets more popular, there will be more and bigger pools, making the
threat less over time.

>
> I think best would be to get the arbitrators down to the user level with very low entry costs, otherwise it could lead to some power concentration like we see it with mining pools in BTC.
As discussed above, low entry costs (both in monetary terms and in
identity terms) I think will not work due to Sybil.

> Maybe some structure like in liquid democracy could be used:
> Everyone has the basic right and possibility to be an arbitrator. If he does not like to take that chance/job he can delegate it further up in a trust line (give his arbitrator right to some he knows or abort it).
>
Such things require a pre-existing trusted identity system.

> br,
> Manfred
>