Gaming farming rewards - cost for GETs

Besides that (I’ll come back to this interesting game) there is something else that will put a cap on both legitimate and illegitimate rewards:

The maximum number of close nodes that a given vault has, together with cache TTL and the effective decrease in cache TTL as a result of churn (which essentially acts as cache eviction).

This is calculable. It’s probably good to do this regardless of any discussions on abuse, since it is very interesting to know where this upper limit to gains are.

(Or just show that there is some error in above logic)

2 Likes

Hold your horses (as I said, I’ll come back to our little game :sweat_smile:), I am stating that it applies to legitimate as well as illegitimate gains.
The purpose of the thought experiment is to

  1. Establish what this cap is (or if it is a logical error)
  2. Get some feedback from designers on how thoughts go around this cap.
  3. Include in general calculations on viability of farming.
1 Like

lol… that reminds me of autosurfing…

In the realm of Internet marketing, autosurfs are traffic exchanges that automatically rotate advertised websites in one’s web browser.

Autosurf - Wikipedia

boosting the visitor counter (or defraud ad-revenues) of the said site.


https://vitalik.ca/general/2019/04/03/collusion.html

3 Likes

Apart from gaming the system - another point is to not create a system that encourages wasting of resources, such as any downloading just to make farmers have greater rewards. Rewarding GETs might lead into downloading just to create GETs. Or have I misunderstood something?

3 Likes

Maybe somehow rewarding cached deliveries and penalising not executed ‘regular get answers’ could incenticise beneficial behaviour :roll_eyes:

The trick now is that it should not increase income to have high frequent gets…

Up up up, credit where credit is due. It was actually @torss who came with that idea :wink:
And besides, now you have given a spec for MaidSafe to work with; “Things the system must handle”. So there won’t be much business on that front for you I’m afraid :smiling_face:

2 Likes

What you think about the solution I proposed here?

We don’t need to reward GETs nor PUTs, not any action, just the potential.

1 Like

Mmm, well, if it’s too complicated it won’t be worth it. Solution must be simple and effective. Otherwise it’s better to skip it.
Now, simple and effective solutions are the hardest to craft, so that does not mean there’s any giving up preemptively.

The idea is not to earn nothing, but earn a reduced farming rate. They still have some value so will be rewarded.

2 Likes

Slight OT.

This ties in with something I raised in other topic. Not really picked up by anyone yet (not showing the logic is defunct, nor proposing solution).
There’s a lot of thinking that goes out into the void now I think.

So, when price is set in every section, depending on the inner state of that section, we will get a variation in price which probably has a normal-ish distribution. This means that there will exist extreme price differences between sections, even though statistically, you will have a very uniform price when interacting with the network.

So, is this
A. Not a concern (why)?
B. A concern technically?
C. A concern PR-wise? (“unpredictable price” seems like a misunderstanding that could stick…)

I think that farming algorithm could be improved significantly if metrics was shared between sections. Can that be done cost efficiently, and without compromise of security and similar?

3 Likes

I like the idea - I’m just not sure how potential can be measured without it being actually used - but they could be tested by the elders I guess - and if performing gets that have been observed in the past and measuring response times even non caching nodes could be identified

2 Likes

I guess it would mostly be a concern if you had a MD/AD that resided in an expensive section. Normally when you upload a sizable file, it’s chunks would be spread to different sections, so you’d get more of an average price. It would be nice to pay an averaged price, but would that further strain the already strained sections? Maybe it could hasten a section merge?

1 Like

Mm, that’s right. And even so, the absolute amount of that can’t get very high since that MD/AD doesn’t actually store much data. (Well, the AD could theoretically store u64 * entry size (said to be 100kb) amount of data, and if it will be designed to always reside in one section regardless of length then you could incur that cost, if section does not level with rest of system before you have uploaded that… Quite an edge case.)
Probably not a technical issue I can see by this.

The metrics-gossip would serve another purpose than averaged price though (getting more OT here): sections could book keep reported and used space, and therefore include % filled in the gossiped metrics.
Including this in farming algo would give a much smoother curve than the binary Full/Not full metric. It would have much higher informational value.

Now, if it is efficient to introduce this gossip is another question. But I think the data can be useful for the sections decision making on many other areas as well.

2 Likes

Great article - thanks for linking.

I didn’t actually see his post, so that is at least 2 people with the same independent idea. I actually thought about it last week, but hadn’t had the motivation to post it.

The point is, it was just a cursory thought. I suspect there are many more ways to game the system with some reliability. Those motivated to work long and hard to do so, with an aptitude for hacking and game theory will lap it up.

We have to be extremely careful with incentives. They have the power to ruin the network if incorrectly aligned. Sure, some rewards are required - such as providing tangible, objective, services to the network. Everything else is risky, especially if subjective.

5 Likes

No, I think that the potential is measured by actual use, the chunks they deliver. The vaults should be equal in the content over long time, given the random distribution of chunks and periodic relocation. So we could as well reward them just by the amount of space (and speed) they provide. I guess it’s kind of similar to bitcoin mining pools. You would be rewarded by the capacity you provide, never mind if in a certain time period it is used more or less.

@oetyng Maybe the price difference of PUTs in different sections could be balanced by the network with the Safecoins it owns? So that if the file ends up in more expensive section the network would pay the amount that exceeds the average. And this would be balanced by the network taking the extra for itself when the price in a section is lower than average? OK, I don’t know at all about how this would be done.

I just so wish that SAFENetwork would not reward any wasting of real world resources, and downloading just to increase the farmer reward would be bad in that sense.

But then PtP is still bringing people into the safe network. It means there is people using safe since people will not come over to the safe network just to do a few gets on a random exploitable app. AND PEOPLE GET BORED with such things. And as you say the abuse is easily abused itself. And like “the pay to read emails” programs, they peak early on then eventually die because the returns are always low for the efforts (even using bots) and people just get Bored too quickly and of course there will be so many that none remain effective

Rewarding cache specifically opens up gaming big time. That means there is nothing to control gaming of farming rewards.

What if every thing was rewarded or penalised perfectly in portion to the effect (every GET, PUT, relocate, parsec vote, split, merge, cache hit, etc etc)? Putting aside whether this is practical, let’s just see if it’s possible. I suspect not, since each of our lists for how value ‘should’ rightly flow would be a little different.

The task is to decide what compromise is acceptable and how to implement those compromises, not to make an ungameable economy.

The ability to ‘game’ the system comes down to who considers what is fair or unfair.

Some rhetorical thought experiments:

Is it wrong for a deviously smart person to be rewarded less than a well-intentioned smart person?

What value do spammers derive from their spam and is it reasonable for the network to object and declare their spam is not valuable?

Can there be any one singular set of correct use cases?

Is growth of any kind desirable or only specific kinds of growth, and why, and how can this be managed?

It’s useful to talk about what’s valuable in the network and which rules will best express those values, but it seems like this thread will struggle to get past ‘measuring sentiment’, despite the apparent desire to be a way to ‘decide the rules’. Very interesting conversations happening here regardless of what ends up actually happening :slight_smile:

I lean toward simple clear economic rules that favour emergent behaviour, a network that’s empowered to manage resources for their health, and a flexible vault-level mechanism to define the fluctuating boundaries of value flow (eg rate limiting / cache expiry policy / fault tolerance levels).

7 Likes

The reason why cache hit cannot be separately rewarded as it is perhaps the easiest to game. It doesn’t require you to have much of a vault or anything like that. All you need is to see that a second machine of yours cached a chunk.

The first machine trawls the datamaps of the public data and sorts all the chunks by XOR address and uses a binary search through the list and when the second machine gets a corresponding cache store then try again and see if it supplies that chunk a second time, then hit that chunk as fast as the second machine can deliver.

One purpose of the cache apart from speeding up the network is that it majorly reduces the effectiveness of gaming farming rewards by requesting the same chunk over and over again

1 Like

I believe @mav suggested payment should only be proportional to effort, which would nullify any gaming attempts.

1 Like