3 ms·

You seem to be right! That doesn't dramatically change the problem, but with avg blocksize 13Kb https://etherscan.io/chart/blocksize https://etherscan.io/chart/

You seem to be right! That doesn't dramatically change the problem, but with avg blocksize 13Kb https://etherscan.io/chart/blocksize https://etherscan.io/chart/blocksize multiply 6000 blocks per day the data to download per week is just 700Mb, am I right? I wonder why the wallet syncs so slowly then. Shouldn't believe random charts on the internet!

CryptoPunk 9y ago
That's 78MB :) Eventually other solutions will be needed, like sharding. One solution for the never ending growth of spent transaction and block header data is snapshots. Practically, you can safely not validate tx data older than X, if X is a sufficient length of time. A solution for neverending state size growth is storage rent with pruning of entries that default on rent payment.
- homakov 9y ago
  78MB per week recently? What's the math? Isn't sharding major compromise on security? I thought snapshots mean you just download the exact state at block X and never bother about tx older than X. And you do that not because of time, but because at the point of installing of blockchains you already trust _someone_ with node software, so you should also trust with latest state (and treat it just like genesis block). Speaking of storage tax, Ethereum should have implemented one like yesterday.
  - CryptoPunk 9y ago
    Apologies, I misread what you wrote. I thought it was 6,000 blocks a week. I also misread that as 13 KB per block header, though that mistake didn't change the calculation. If the average block size is 13 KB, then there should be 546 MB added per week (13 KB / block * 6000 blocks / day * 7 days / week). I don't know enough about sharding to offer an informed opinion on it, though I imagine the security compromises are tolerable. With regard to a snapshot, yes that is one way they're used. I was suggesting that archival nodes could even rely on snapshots to cap their storage requirements.
    - homakov 9y ago
      It would be perfect to make a snapshot whenever a blockchain to last state is taking over 20% of state size. I.e. with 10 GB state do it every 2Gb (monthly). And IMO it's safe to discard any chain older than 3-6 months because only people who were offline for that long would ever need them (unrealistic)

CryptoPunk 9y ago

That's 78MB :) Eventually other solutions will be needed, like sharding. One solution for the never ending growth of spent transaction and block header data is snapshots. Practically, you can safely not validate tx data older than X, if X is a sufficient length of time. A solution for neverending state size growth is storage rent with pruning of entries that default on rent payment.

homakov 9y ago
78MB per week recently? What's the math? Isn't sharding major compromise on security? I thought snapshots mean you just download the exact state at block X and never bother about tx older than X. And you do that not because of time, but because at the point of installing of blockchains you already trust _someone_ with node software, so you should also trust with latest state (and treat it just like genesis block). Speaking of storage tax, Ethereum should have implemented one like yesterday.
- CryptoPunk 9y ago
  Apologies, I misread what you wrote. I thought it was 6,000 blocks a week. I also misread that as 13 KB per block header, though that mistake didn't change the calculation. If the average block size is 13 KB, then there should be 546 MB added per week (13 KB / block * 6000 blocks / day * 7 days / week). I don't know enough about sharding to offer an informed opinion on it, though I imagine the security compromises are tolerable. With regard to a snapshot, yes that is one way they're used. I was suggesting that archival nodes could even rely on snapshots to cap their storage requirements.
  - homakov 9y ago
    It would be perfect to make a snapshot whenever a blockchain to last state is taking over 20% of state size. I.e. with 10 GB state do it every 2Gb (monthly). And IMO it's safe to discard any chain older than 3-6 months because only people who were offline for that long would ever need them (unrealistic)