No I meant it as
- today it might cost (all inclusive cost) x amount to incrementally handle one chunk
- Next year it is likely (averaging every farmer obviously) to be 1/2 x
- year after 1/4 x
Now the 1/2 might be 40% of x or 60% of x and vary slightly from year to year. And next year or two might see a 10 times reduction in the cost of SSD storage, so sometimes the reduction could even be much greater.
Now I didn’t include the statistic that says for an average file that as it gets older it is accessed a lot less year by year. Now this isn’t for OS files but files people store. Think of those dvd movies you or someone you know has bought. The first 3 months you might watch it a few times, then the next 9 months you watch it again less than you did in the first 3 months. Then the second year you watch it a lot less if at all.
The point being that averaged across all chunks the older the chunk is the much less its read. This impacts bandwidth in a way that sees new data consuming most of the usually free bandwidth costs. So even for those who pay incrementally for bandwidth those cost dramatically reduce for a particular chunk over time.
So if most farmers due to economics are on unlimited (or extremely high quotas) and their bandwidth cost is insignificant or zero and only a few are paying incrementally then dramatically reducing the very small overall bandwidth cost for a chunk over time is extremely important. Important in the sense that rental needs to charge 80% of the normal cost up front just to cover the the statistically shown usage of the upload for bandwidth portion of the costs. In other words its just not worth the effort and the complexity of a rental system.
Also if you look at industry and the consumer rental system for electrical/electronics, you see that the actual rental charged ends up so much more than paying up front. And due to the complexity of introducing rental into the code will cause a lot more processing and bandwidth and disk activity that the actual cost to store rental data could even be twice the cost of pay once data. Pay once sees disk activity to just store and whenever its retrieved. Rental sees daily checks if the data is to expire (daily disk activity for each chunk) and bandwidth usage for every node in the section to agree on whether to delete or not.
The point is that rental has its overheads that must be charged to the renter. For the computer industry it is like 40% break even then rental firm charges a profit on top. I think from a guesstimate that the network would have to be charging something similar for the first couple of years.
Rental is a can of worms (coding, user experience, costs) that really makes it a cure that is worse than the perceived problem.
I honestly think you way overestimate this cost. By a lot.
Anyhow farmers are paid to retrieve data so that covers those costs even if they are small (or zero for most)