New Proposal to reduce space requirements for full nodes

[BigG]

We cannot expect thousands of full nodes to store any transaction ever occurred. It has to be sufficient to only save the transaction origins not older than two years.

There are around 371k Blocks so far. Let's compress the first 200k to only wallet balances in one file. When we reach 400k, the next 100k blocks are compressed and then deleted.
So we create every 100k a new master-block out of it and still have the transaction origins of at least 100k Blocks or almost two years.

Would that less dynamical “pruning” compared to my first proposal work? Any security issues I miss again?

[1538090185]

1538090185

[1538090185]

1538090185

[1538090185]

1538090185

[jl2012]

Please read section 7 of Sathoshi's paper: https://bitcoin.org/bitcoin.pdf. Bitcoin Core 0.11.0 supports pruning already.

[goatpig]

I am aware of that. But today you cannot spend coins which are in the pruned blocks. That should be solved.

Why not? My understanding of pruned blocks is that they keep all UTXOs with the necessary intermediary hashes to compute the merkle root. What pruned blocks get rid of is transaction history, unspent coins scripts are still all there.

[jl2012]

I am aware of that. But today you cannot spend coins which are in the pruned blocks. That should be solved.

This is just the problem of the 0.11 implementation. Not a protocol issue. They will fix it in future version

[gmaxwell]

This is just the problem of the 0.11 implementation. Not a protocol issue. They will fix it in future version

Not even a problem-- the wallet support was just untested so it was disabled to avoid delaying the feature. It more or less worked from the very first patch (as 0.8 radically change the design of the software to accomidate that-- we've been working towards it for a long time).

[DannyHamilton]

Would it be possible to create a hash out of the first 100k Blocks and declare that as a new Genesis-Block containing all the relevant wallet-balances?

Even if this was possible, and even if you could get someone to implement it for you...

Why bother?

The total size of the first 100,000 blocks is only about 10 megabytes.

You'd be shrinking a 47.5 gigabyte blockchain all the way down to 47.49 gigabytes.  That certainly doesn't seem worth it.

[mezzomix]

It is worth to think about "forgetting transactions origins" after a few years.

This problem is already solved with the pruning implementation.

Then the blocksize does not matter anymore.

The problem with pruning is that you can't bootstrap a pruned node without downloading the complete blockchain. You do not need to store the blockchain but you need to download it once to create a current UTXO database.