Current farming reward distribution will more likely promote centralization rather than reduce it

I was reading a bit through the docs, and came across the section about preventing farming centralization. Apparently the farming reward does not scale linearly with the resources a farmer provides to the network. Instead it uses a sigmoid curve which penalizes bigger and smaller nodes This is supposed to prevent centralization in farming nodes.

I want to argue that this actually will not reduce centralization, rather it will promote duplicate accounts/nodes and if it has any effect on centralization it will more likely increase centralization rather than decrease it, depending on the measures taken to restrict account creation.

Lets say my hardware provides 10x the network average of resources, but only gives me 3x the reward, whats my incentive to not run 10 nodes, each providing the network average of resources, giving me 10x the average reward?

First there is a fee on account creation, this will have little effect as it merely adds a one time investment which will be paid back by the increased farming yield over time.

Then there can be technical restrictions at the protocol level: one node pr ip, some kind of hardware signature. I have even seen discussion on proof of uniqueness at the protocol level.
So lets say we put in some mechanism that makes it more difficult to run several nodes on one computer, or on one ip address. The what and how is not so relevant here, my assumption is that there will always be a way to get around any such mechanism, and at the same time they will put restrictions on legitimate uses of new accounts.

So what will happen here from an economic perspective: Lets say i have the means (time & money) to research an efficient way to spoof the mechanism and invest time in setting up my network/computers that allows me run the optimal amount of nodes to give me the highest possible farming reward.
I have basically made a one time investment which allows me to get a higher yield on the resources i provide compared to a regular users with above average amount of farming capacity. And the higher the difficulty of spoofing will only cause a smaller group to get a better advantage, basically promoting economies of scale.

So to sum it up, the sigmoid function by itself will only promote “duplicate” accounts which gives us a false view of the actual network topology (100 nodes on one computer).
Adding restrictions to account creation will merely benefit those with the means to get around these restrictions, while imposing restriction on possible unforeseen legitimate uses of multiple accounts.

would like to hear your thoughts. and to clarify, im not so worried about this as an attack vector, only how it skews the economy and adds unnecessary complexity with no real gain.

Yes, that (the reward curve) is a workaround that cannot prevent centralization because there’s no way to tell one human from another. Even if you could read the machine ID you wouldn’t know who owns it (and in theory it’s possible - although unlikely- that two people such as roommates run separate Safe VMs on the same machine).

I wrote 1-2 posts on centralization here before so I won’t repeat what I said there, I just want to add that botmasters may find it profitable to install and run Safe (in addition to all the other stuff) on computers of unsuspected victims because anyone with free space has the potential to generate additional revenue for them.


I don’t understand your reference to “spoofing.”

Running multiple nodes on a larger machine is actually recommended, as opposed to running one large one. This is not spoofing. It serves the network to have more nodes, regardless of who owns them or how many machines they are running on.

If a lot of resource is being contributed, then a lot of reward will be available.

But remember that adding huge amounts of resource, way beyond what the network needs, will just result in less productive (less profitable) farming. It will actually result in a more efficient network, though. I think farming is designed in a way that nobody will be getting rich or even making a living on it. It should balance as an equitable swap of resources for resources, with a slight added value due to safecoin’s utility for speed and anonymity of exchange.

Yes, there’s nothing to stop persons or companies that have lots of time, money and resources to throw at the network from getting a sizable amount of the safecoin produced, but in that case they ARE providing a lot of resource, which is good for the network. Smaller users will still earn SOME safecoin, and the network will be faster and more efficient due to the large amount of resource.

There is a potential attack vector here from malicious actors who have lots and lots of server idol server farms who just want to create mischief. Bringing large amounts of resources on and off line simultaneously might cause huge churn to be a problem, but that’s another topic.


Now that is an interesting possibility, because it forces another to contribute the recourse to the benefit of the crook. I don’t know about the capabilities of such botmasters.

as in setting up on big server farm disguised as many smaller nodes. If the purpose of the sigmoid distribution is to prevent centralisation, which the article clearly states it seems very flawed.

So if i understand you correctly the sigmoid function has more of a address-space balancing function (if that makes any sense). Or at least some form of technical optimisation of the network?

In that case it should at least be described that way, and hopefully also implemented that way (ie. the amount of vaults is handled by the client in the background)

Totally agree, i thought the sigmoid was meant to be a roadblock to prevent that to some extent.


Adding 100 vaults to the network doesn’t give the person doing it any more control over the network than the person adding just one. Control is the keynote of centralization.

The scheme does combat centralization to the degree that it levels the playing field of what devices can expect to get in relation to each other. If it was more profitable to run a vault with a 10 terabyte drive, those are what would flood the network from “richer” sources and crowd out others altogether. That would tend to squeeze the smaller participants out altogether.

I think the curve also provides a proper balance between processing power and bandwidth in relation to storage. That makes message handling, consensus configuration and all the other things the network participants must do to be a viable vault much more effective. A small processor with low bandwidth that accumulates 1 tb of storage is likely to choke more often than ten 100gb vaults with similar processing. The more balanced ones will contribute more to the network and, theoretically, be rewarded better.

I think that’s the balance that is trying to be struck.


And if it has not already been said. It is way more profitable for the home user to add vaults from their PC that they are using. It costs next to nothing in electricity, h/w costs and the framing profits are basically all gravy. But the centralised framer does not have this luxury and all h/w costs and electricity reduces any income from framing. Also the algorithm really means that the home framer is expected to make a reasonable profit/ROI and not a living

As @fergish said, once available (unused) space grows the farming reward drops to almost nothing. This means that as centralised farming grows it will become exceeding unprofitable. If it grows too much then there would be only 1 coin attempt for 2^64 gets. That is untenable for commercial farming but just unfortunate for the home user, that is the commercial farmer will move on to something else and only a %age of the home users will stop, until enough stop and the home user gets coin again.


It will be interesting to see what will be considered “commercial”.
At the moment some folks plan to buy a semi-dedicated home minifarm.
We don’t consider that to be pro-farming, but if enough users do it, it may turn to be untenable in terms of ROI, in which case using spare capacity on existing hardware (i.e. essentially zero cost farming) may turn out to be the only non money-losing approach.

In post below I claimed that even RAID-ed setups will not be viable if dedicated to SAFE. Because of that I’ve always claimed that questions about “best” RAID level are moot. The best level will be no RAID and I believe discussions related to RAID will have been a waste of time.



I guess I’m your guy for “commercial”. I plan to use all of the excess resources Klouds amasses as a farm. My intention: provide SAFE with well-peered, high-quality network access globally. Klouds main target isn’t individuals: It is instead “legacy hosting providers” who have yet to adapt to the new era.

Deopending on payouts and the like, I’m ready willing and able to rack up hundreds or thousands of nodes on GCE or AWS. Their preemptible/spot instances make this relatively affordable.

My theory: We’re not going to know what the scarce resource is until it is widely in demand. Here are some of the possiblities:

  • Storage
  • Compute
  • RAM
  • Well-placed nodes in terms of network topology (everyone needs this. I assume that this will be one of the most in demand items.)

The usage of the network is going to shape the scarce resources, and I highly doubt that they’ve gone and built a static payment model that is unresponsive to the needs of the network.

1 Like