More Economics (15)

Almost pure PUT: backups
Almost pure GET: netflix

It’s definitely possible to have a strong bias toward one or the other. It doesn’t depend on a functioning market for resources, it depends on consumer behaviour. Or to put it another way, resources depend on consumers, not the other way round.

In reality the difference should hopefully end up balanced across the different use cases, but that can’t be assumed. For instance, bitcoin was assumed to be both a payment solution and store of value, but it ended up being very strongly biased to one side of that.


In ancient ADSL lines, asymmetric speeds with low upload latencies was common but modern optical fiber lines are mostly symmetric and with quite good latencies and can compete easily with datacenters that, although have better connections, should serve many users at the same time. And in mobile, 4G and future 5G are not symmetric but close with a ratio 1:2.


I agree with this - discussion at a high level about what the aims are and requirements could be more productive until there’s some framework of common understanding to base discussion around (including a base of common assumptions, which we may well not have right now, making discussion difficult / frustrating).

I’d like to work on something like this myself, but unfortunately don’t have the time just now. Maybe in a few weeks time!

In the spirit of this, I won’t go through all the points or keep discussing without a framework to refer to, but will respond to a few points briefly:

I’d call it a market failure if the network were charging users more than required for farmers to provide a given amount of resource. It would make using the network more expensive than necessary, which shouldn’t be desirable.

Can you give an example of how the network could have a problem keeping all data safe and secure if it balances supply and demand around a target level of spare resources (e.g. 30% spare resources)?

When would the network ever need to increase spare resources over the optimal target level if there’s no increase in demand from users?

When would the network ever need to reduce spare resources to less than the optimal target level if there’s no pressure on resources?

I feel the ideal size for a network currency buffer is Zero. The only buffer required is the spare resources on the network (e.g.30%).

There will be a constant stream of income from PUTs, and a constant stream of payments to farmers. The network merely needs to balance the pricing / reward levels so that the target resource level is achieved.

Unless there’s a practical reason for a buffer, why have one?

No special mechanism / buffer / purse is needed for smoothing because the ‘sensor’ that is used to determine price changes (spare network resource level) is sufficiently dull to ignore transient peaks and troughs, but sufficiently sensitive to medium term trends to take price action ahead of any issue of significant over/under supply.

This all needs discussing and formalising theoretically well ahead of designing and finalising algorithms. The algorithms should be designed to follow sound and detailed economic reasoning, not without it & hoping for the best.

Once a market is established, prices should remain pretty stable because they’ll be set around the cost of farmers providing resources to the network, which will be quite stable (though hard drive price spikes might be felt etc).

A large, active initial purse for the network could do more to destabilise prices for farmers and users if it reduces the effectiveness of price signalling in the market.

You assume that a large initial purse can help these things, but I can’t see any market mechanisms that would require a purse in order to achieve stability.

@Deadloch has made a good start on setting out some assumptions etc, perhaps I’ll just try to build on these when(if!) I have some time to help build (or at least work toward) an agreed basis for discussion about the SAFE network economics (e.g. simple supply / demand diagrams for various aspects, lists of assumptions, levers, and sensors, common terminology etc).


I think it’s fair to assume that if the SAFE network is useful, there will be demand both to put data onto it, and download data from it. From day one people will be doing at least the concepts mentioned on this forum; music sharing, video sharing, blogs, forums, web sites, chat, email, data back up, etc, many of which need PUTs and GETs to work.

If Netflix created their own version of the SAFE network which only they could upload to, they would need to tweak the economic model to suit, and bandwidth would likely be the major constraint on farmers, not storage.

As you say though, on the SAFE network, there could be a strong bias towards PUTs or GETs. This shouldn’t be a problem for a market balancing supply and demand without any financial buffer, as long as there is a suitable mechanism of rewarding farmers in place that makes provision for this eventuality (i.e. sufficient granularity of farming rewards in the case of relatively few GETs).

One potential issue is that, while a strongly PUT biased network would have a direct pricing lever to ensure farmers increase resources at the required rate & demand doesn’t exceed that rate for long (by increasing farming rewards & PUT prices), a strongly GET biased network wouldn’t have such a direct pricing lever to tell farmers it needs bandwidth rather than space. Farmers would likely figure out the bottle neck if small vaults resulted in massive bandwidth, or if farmers with faster connections were receiving significantly greater farming rewards than those with slower connections, but it might be slower for farmers to add more bandwidth than more storage space.

I agree. This table of Network Health Metrics may be useful. Not to say it’s official or definitive or whatever but it’s a start to building a common language and understanding that hopefully leads to a clear design spec for the safecoin / security / incentive algorithms.


I think we also have to note that data in high GET demand is likely very useful. If the cost to PUT it is way lower, we should consider what value per PUT that has provided. Not value as in money, but value as in usefulness.

Ultimately, we are likely to have a lot of stuff put on the network which is of limited value. These will subsidise the good stuff and that seems like a nice place to be to me! :sunglasses:

1 Like

Interesting thoughts. I guess it’s hard to judge the usefulness of a PUT only on how many times it’s downloaded. There could be files that are highly valuable that aren’t downloaded much (e.g. encryption key backup), compared to a mildly valuable file that lots of people want to download (e.g. cat video).

Thinking along these lines, I’ll be surprised if we do end up with a bandwidth-constrained network rather than storage- constrained, because while lots of people will want to store far more than they can fit on their hard drive (e.g. backing up photos / videos), they cannot easily use more bandwidth than they can use at a single location. Storage is also taking up resources constantly, but bandwidth only when actively downloading / uploading stuff.

If this turns out to be correct, the network should be storage constrained, so it won’t really matter how useful data is. Even if lots of data is very useful, the price of storing more will only reflect he cost of farmers adding additional storage, and it’s the same for less useful data. I guess less bandwidth demand will make farming less annoying for non-dedicated farming machines, so more people will be willing to dedicate more resources to farming.

1 Like