I don't fully understand why PoI is sustainable

Technical Discussion
#1

while I was translating NEM’s technical reference into Japanese(http://nemmanual.net/NEM_Technical_reference_JA/), I noticed that PoI seems to have centralization effect severer than other PoS schemes.
The problem is, PoI has no means to punish spamming(i.e. Transacting XEM between accounts controled by single entity.). What it can do is to only set “upper limit” for those malicious transaction by NCDawarerank.
In technical reference, it says that this problem doesn’t matter because …

Because PoI scores sum to unity, other rational actors should take the same actions. This
would cause the expected benefit to disappear

But the truth is, it does. because

  1. the transaction fee will have more severe effect for who doesn’t have XEM than who has.
  2. who have more XEM will be more probable to recollect their transaction fee since they can run super-node.

And if it doesn’t, it will cause higher network consumption than PoS.
This will cause centralization effect in long run and therefore NEM will collapse.

So I suppose, NEM needs something like Proof of Individuality (http://proofofindividuality.tk/) or something similar but more loose.
Are devs facing to this problem while building CATAPULT? If not, why?

4 Likes
#2

Honestly, I don’t see the problem.
You are aware that tests have been performed and that there was even a challenge where lots of community members tried to boost their PoI score in the early days. In the end the conclusion was that you can boost your PoI score but at the cost of transaction fees ( which are far higher than the benefits).
Spamming is punished due to the fees involved.

  1. I don’t get what you mean here but if you don’t have XEM , you won’t have PoI …
    2.This is irrelevant to PoI

I’m not sure whether you really understand how PoI works, but I can’t figure out what the issue is.
Still, I don’t see the flaw. If you really think there is an issue, I’d like you to explain how you would exploit it…

1 Like
#3

when the network might be heavy used and the possible to harvest fees are much much much more as now (like 100-1000 time more) then maybe harvest on a node that is in center of activity can give ya a income boost

if that are real transaction or transactions to boost harvesting importance might be nearly imposible to dedect

but it even doenst matter

there is no fee free transaction in NEM and the more transactions the more fees can be collected

i see no issue

even if some transactions just initiated by specualte on raising own node importance and create more blocks

in fact u raise the harvest income for everyone who harvest by fill the network with transactions

2 Likes
#4
  1. I don’t get what you mean here but if you don’t have XEM , you won’t have PoI …

I mean “less” XEM, not “no” XEM.

#5

Aaaaaaagh!!
OK you’re right, I don’t understand PoI,

what makes me nervous is that the technical reference is good enough for explaining “how” it works to some degree but not for “why” it has been defined to work in that way. for example

  1. why transfer fee includes arctan function? not log ?
  2. how did you select parameters for NCDawareRank (i.e. η, and μ) and for PoI itself(w_o and w_i) ?
  3. In “7.6 Resistance to Manipulation”, why it only shows that boosting the importance is possible but doesn’t show that the benefit for boosting Importance will be offset by fee? And most importantly, why it had to say

Some actors can boost their importance by splitting their accounts or sending transactions to other accounts to mimic economic activity. Because PoI scores sum to unity, other rational actors should take the same actions. This would cause the expected benefit to disappear.

if it has been tested right after launch that there are no benefit ? (By the way, you have done lowering the fee by hardfork after this test, right? than this does mean nothing unless you have proven that the new fee won’t cause problem)

Another thing, Is there a source code for simulation or something that would be educative for me ?

#6

Those are inherent consequences of the economy I guess. For those with more money everything is proportionally cheaper than for those who have less. You’d be showing lots of people a favor if you find a way around that :wink:

Further, I can’t answer most of your questions since I was not the one who designed PoI but I can only guess that a theoretical model was created and then tested to fine tune the parameters.

Most people don’t, but it has proven itself to be useful :slight_smile:

Anyway, maybe you should edit the title to something less alarming :wink:
People without the proper knowledge might start questioning PoI based on things like this.

Lower fees means lower rewards from harvesting, that too is a zero-sum game so I wouldn’t see how this would affect the validity of PoI.

2 Likes
#7

the goal of the tests was to prove that a active address have more network importantce and more harvesting income
but only if its naturaö traffic means transactions that u send and recieve anyway

because then u didnt invest the tx fee to harvest more u had to pay them anyway

if u create the tx only to try harvest more coins then the test did show u spend more xem in tx fees that u able earn more from harvesting

mean goal that naturall activity is rewarded and artifical tx activity not

2 Likes
#8

You are right … this reminds me that PoW has succeeded in that way.

I think I’m starting to get the point. This is either of

  1. Technical reference is wrong. There is absolutely no point for spamming. (because of high fee compared to harvest rewards under any circumstances.)
  2. There could be benefit for spamming, but only when there are a lots of honest transactions in NEM network. This won’t be always true so it’s negligible. And it is still better than PoS regarding to the centralization risk because benefit of cheating should be low(I still think this requires quantitative argument though).

Am I making sense? If I am, which is correct, 1 or 2 ?

1 Like
#9

Generally, POI was thought as a way for smaller accounts (let’s say accounts not owning millions of xem) to additionally gain some importance by making transactions. I am not sure how many people actually have tried it out in detail since the harvesting fees are small and transactions cost money too.

Big accounts cannot boost their POI big times since the bigger the account, the more POI behaves like POS. That way we ensure that the algorithm can’t be badly tricked.

So while a small account with 100k xem can boost its POI by a factor 3, a big account with 100M xem won’t be able to do that.The latter can gain maybe a few percent by transacting, so nothing which we should wory about.

3 Likes
Алгоритмы, используемые в NEM
#10

Thanks for your reply, but honestly, I’m still not comfortable with the explanation I’ve got so far.
@BloodyRookie is talking about “account”, but I am talking about “actor” who is in behind those accounts since that is what really matters in terms of cryptocurrency.
actor can split his account to small accounts, which has only 10000XEM and therefore boost his importance. or he can send XEM in between his accounts if he wishes to.

It could be true that this will have slight benefit because of NCDawareRank. But, OK Please listen, the most important point I want to make here is that actor who has more XEM has more choice than the actor who hasn’t. This seems a big issue to me, since if an actor behaved to maximize their benefit, this makes PoI (at very least) as same centralization effect as PoS.

Yes, “so far” it has been proved itself to be useful. And this could be because most people really don’t understand PoI.

(I’m starting to think that I should change the title of this discussion as before…)

1 Like
#11

just splitting an account won’t boost importance and just transacting between those accounts won’t boost the importance too.If you do the latter then those account form a cluster and the importance of the entire cluster will stay the same no matter how many transactions you do within the cluster.

Not sure what you are aiming at. Rich people always have more choices than poor people, that will always be true.
POI is no magic concept turning poor people into powerful people.

3 Likes
#12

I have done lots of testing and I have found that making many small accounts that are active increases my POI versus having one large account. That said, it quickly becomes complicated to manage that many accounts, and while I will get more fees from blocks, I also have to pay some fees to split those accounts, and pay fees when transacting from them, and initially lose POI as that transferred XEM has to be aged in over 30 days. So while it is possible for a rich person to get an additional boost by trying to play the system, it is not free, it isn’t that easy, takes a while, and carries some additional risk as they will have more accounts to manage. And ultimately, even if a rich actor set up a series of bots that maximized and automated all this, they still wouldn’t gain overly excessive gains. As a whole, consensus and network security would still work quite well.

1 Like
#13

@jabo38 is making the point. what I really want to be sure is that if this

is really true(even if there are much fee to harvest). I want to know the reason you think like this. since its not clear from the definition of PoI(at least to me, and probably for many people who are hesitating to invest on NEM).

Yes, but that choice shouldn’t be the choice for increasing the amount of harvested XEM, at least compared to PoS, since that is what PoI meant to be.

2 Likes
#14

Hypothetical question:
If somepoint someone is able to build somekind POI money machine, or some new idea’s comes even without problems, can POI system be fixed/improved?

#15

Let me sophisticate my answer a bit more.

@jabo38 is saying playing the system is not free. This is true but the income for playing it will raise if there are much transaction taking place in the whole NEM ecosystem(this will be true if NEM has gained popularity.) and

This is not a risk at all. Building a bot is quite easy if you can define the optimal strategy. And consensus algorithm generally has to be built under an assumption that people are trying to maximize their own income.

I think this is possible by hardforking. What we are arguing here is not about the vulnerability which will cause the collapse at one day. So there will be plenty of time to test the new system.

#16

OK some people might still don’t understand what I am arguing here so I will give you more detailed example.
imagine three actors.

  1. Vendor
  2. Customer
  3. the Cloud(whole other NEM economy.)

Vendor has 3 accounts, and each of them holds 10000 vested XEM. He is trying to sell some product to Customer by 3000XEM.
(1) The vendor can tell the customer to pay 1000XEM foreach account by assuring that he(the Vender) will pay slightly bit more than additional fee.(This will be done by discount of the product in reality)
This will cause the Vender to have slightly more Importance than receiving by one account.
Everyone else in this topic is saying that this wont pay since the fee will offset the increase in harvested XEM. But that depends on how much transaction is taking place in “the Cloud”. so there are some situations which will pay, especially after NEM has gained popularity.

And another strategy for the vendor is to (2) send transaction in among she’s own accounts. This will also pay if there are enough fees that she can harvest.

These two strategies makes thing even worse if she combines these together since she can control which cluster she’s accounts should join.

The point is, she can take these actions only when it pays and otherwise she will not take the trouble to consume she’s fee meaninglessly.

you may say "everyone else can take the same action so there are no additional capital gain."
But this is wrong because

  1. it will consume network resource (it will bloat it in fact, since spamming will be more profitable strategy if there are lots of transaction in the first place.)
  2. not everyone can take the optimal strategy, especially ordinary users who are using a mobile wallet.
  3. even under the assumption that everybody is taking optimal strategy to gain his own income, it will still cause centralization effect because if you get richer and richer, there are more strategy you can take. (think what if the Vender has only 10000XEM at the first place?)

You may thought "wait, but there has to be a lot of transaction in ‘the Cloud’ to ‘playing’ get profitable, so the total Importance of the cloud will be much higher than the Vendor’s and there are no use for spamming."
This is wrong because in reality, there are no such a thing “the Cloud”, there are only actors who are playing the same game. some of which have more choice, and some of which don’t.

In one word, PoI is not decentralized at all, It’s even worse than PoS if we assume everyone understands the rule and trying to maximize their own income. (with one exception… if there are absolutely no transaction in the whole ecosystem, it will be exactly the same with PoS)

I don’t think this is a dead-end for NEM, not even for PoI, since I think I can come up with solution.
But I am pretty sure that NEM won’t succeed as cryptocurrency if devs ignore this problem.

Am I making sense here ?

1 Like
#17

I think you’ve missed two CRUCIAL points from the NEM whitepaper:

  • balance vesting
  • and recalculation that happens every 359 blocks

(plus one additional: you know that all the importance sum to 1. right?)

Actually I have re-read all your posts in this thread, I think you’re missing few more things:

  • cluster detection - the point of cluster detection is to limit boost of PoI if transactions are made within a clique

and one more thing, you seem to be assuming that attacking person will generate every block… o0
obviously (s)he won’t, and that means that resources (fees) (s)he invested to make transactions are lost

3 Likes
#18

Sorry, to interrupt here, but in the meantime i made about 50000 XEM by delegated harvesting. Just waiting for 1 Xem = 1 USD. I´ll buy a Lambo then…

1 Like
#19

clustering prevents spamming transactions within your own accounts as you suggest

supernodes can harvest like any other account with vested XEM.
They are paid from the SuperNode Fund, so they have no advantage of harvesting transactions.

#20

Dang, another 287 XEMs. My record is 1200 XEM in one block !
Is it my Internet-Connection ? Thinking about giving up my day job :slight_smile:
Guys, activate your “delegated harvesting”, you´re missing out, and you help the network. Besides, its like playing lottery for free every day . . . Its fun too. Damn, I have to check my account now, again.
But i will spend it on a good cause, promised.

1 Like