[Lightning-dev] A proposal for up-front payments.

[Bastien TEINTURIER]

While I agree with most of your points, I think there are subtleties to
explore before
completely rejecting the idea.

every use of proof-of-work today (other than to power Bitcoin itself, as
> Bitcoin cannot support itself) can instead be done by using Bitcoins to
> impose this economic cost.
>

That is philosophically true, but the complexity of integrating that small
PoW into Lightning
is much lower than the complexity of integrating **fair, un-gameable**
upfront payments.
And not all PoW is born equal: there are a lot of PoW schemes that have
different trade-offs
than Bitcoin mining (think ASIC-resistance such as variants of Cuckoo
Cycle).

Another key point is that creating ASICs for this PoW is fundamentally
different from creating
ASICs for mining a crypto-currency. Solving this PoW doesn't earn you any
money: it merely
allows you to spam to temporarily disrupt the network.
Since this PoW isn't used in any consensus, we can change the spam PoW
algorithm anytime
we want, making all previous ASICs obsolete.
So it's not obvious to me that anyone would find it viable to invest in
creating such ASICs.

As hardware specialization for the specific Lightning-Network-proof-of-work
> arises, we will find that to practically limit spam, intermediate nodes
> will have to increase and increase the threshold for accepting
> proof-of-work, as spammers are going to switch to the more-specialized
> hardware.
>

That's where I think it can be more subtle than what you describe (I may be
wrong though as
predicting future behavior is hard).

Since I'm ruling out ASICs, we're only dealing with "normal" hardware
bottlenecks (cpu/ram).
That means attackers are not playing at a completely different scale than
normal users.
The cost for attackers to generate an amount of spam mimicking N normal
users will then be
somewhat linear in N (to be investigated further).
That's exactly the same result as upfront payments, where an attacker can
still spam like
he's N users if he's ready to pay a cost linear in N.

I'm slightly playing devil's advocate for the PoW proposal because I think
it's worth exploring
more, even if we eventually abandon it. Maybe you're right and it won't be
as effective to
fight spam as upfront payments: but right now with the arguments I've seen
on this thread,
I'm not yet convinced of that.

Cheers,
Bastien
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/lightning-dev/attachments/20191122/32cc41a8/attachment.html>