Hacker News new | ask | show | jobs
by cwkoss 1813 days ago
> With Arweave you pay a small upfront fee to have the network store your file forever.

How are the economics of this sustainable? "Forever" hosting for a small upfront fee seems like it must be a lie.

Can you only upload small files, so you end up paying 1000x+ S3 prices, with the hope that Moore's law outpaces cost of keeping the file online?

Are reads monetized, so unpopular files will inevitably lose the forever guarantee?

2 comments

Seems like distributed file systems have a 'choice triangle': permanent, cheap or "doesn't allow hosting child porn" - choose one.
the fee pays into an endowment. the endowment is only paid out to miners if the block reward in USD terms is too low to justify storing the full weave.
So, if block reward value grows more slowly than the stored data, the endowment gets drained and when it hits zero the system implodes?

Seems like an obfuscation of the economic problem that doesn't solve it.

the size of the fee is a function of 1) the amount of data you wish to store and 2) the estimated cost of storing the weave between the current time and the end of the storage period ("forever" is actually assumed to be about 200 years for these purposes).

If you were to store 100 TB on the weave tomorrow (the weave is currently 10TB), the block reward would remain the same, but the endowment payout would trigger much sooner.

The endowment fees are sized with the assumption that the endowment will have to pay out immediately and until the end of the 200 year period.

Ah, that makes sense. I think their 'sales pitch' would be more compelling to more technical users (which I assume is the main initial demographic) if the 'forever's had asterisks and the 200+ year assumption was prominently displayed on landing page.

That shifts my expectations from "economically infeasible lie" to "small fee may not be so small, but feasible with proper stewardship and valuable for certain use cases."

In the crypto/DeFi space, superlative marketing copy is more likely to be interpreted as 'potential scam' than other domains IMO.