BIP 33 - Stratized Nodes

Hi,

Please check out my proposal,

https://en.bitcoin.it/wiki/BIP_0033

I want to use the existing Bitcoin protocol to provide this functionality in order to maintain compatibility. This proposal does not affect current Bitcoin clients, but allows the parallel system to operate alongside and sometimes intersect with the main Bitcoin network in a positive way.

Thanks for getting this started.

With regards to the specific proposal, I don't believe it's the best option and still plan to eventually implement the original design outlined more than a year ago in this thread:

  https://bitcointalk.org/index.php?topic=7972.msg116285#msg116285

Namely that you use a new protocol command to set a Bloom filter on a connection. Only transactions matching that filter will appear in relayed inventory. Blocks that are requested will arrive as a header plus transaction/merkle branch pairs. Clients are expected to maintain and track the block chain as per usual, but instead of downloading the whole chain and then dropping the irrelevant transactions, that filtering is done server side. By strengthening or weakening the Bloom filters you can choose your preferred point on the privacy/bandwidth-usage spectrum. It is a fairly simple change to the Satoshi and BitcoinJ codebases but still allows clients to gain confidence in their balance by examining the chain, and this is true even in the presence of a hijacked internet connection (you can't trust pending transactions that way, but you can still trust confirmed transactions).

The filters would be applied to each data block in each script rather than having a specific knowledge of addresses. In this way you can select for things like multisig outputs or outputs which don't use addresses / pubkeys to authenticate.

I could write a BIP for this alternative protocol if somebody else wants to implement it. I was going to wait until I had time to do both BIP and implementation, but I think some simple optimizations to BitcoinJ can keep its performance good enough for the short term.

Thanks for this, Amir.

My initial reactions:

1) This is cool and useful (but see 3) 2) This is significantly less secure than validating an entire blockchain; it's certainly worth working out some use cases here in more detail than just a sample conversation. More on this below 3) What about discovery? Will a client now have the chance to look for NODE_STRATIZED clients on IRC? How do you envision a stratized server decides which transactions to relay/store? Or is it just a caching layer in front of a high quality blockchain service? If it is just a caching service, the question of cache hits / misses is an interesting one as well. 4) What are the economic motivations to run a stratized server? Other than cheating people of course. 5) Seems like a 'send me everything for this source address' is going to save a lot of roundtrip conversations for what I imagine the most common request will be.

Inre: majority agreement on transactions, and even balances, it would be nice to work out some theoretical security / cost / value calculations for the following scenarios:

Likely value and cost to someone of subverting / lying about 1) An n-confirmation transaction, n > 0 2) A 0 confirmation transaction 3) A NODE_STRATIZED transaction chain for a client with m connections to NODE_STRATIZED servers 4) An address balance request for a client with m connections to NODE_BALANCE_INFO (I made this name up) servers

Peter

A bloom filter seems like an interesting idea. However this proposal is concerned mainly with the initialisation stage, whereas this bloom filter is for pushed blocks.

This proposal still updates new transactions and blocks in the same way, and it's not inconceivable to later use a bloom filter to make that part more efficient (although it's questionable if pushing this server side would be a good idea as it would now need to track an additional client state).

________________________________ From: Mike Hearn <mike@plan99•net> To: Amir Taaki <zgenjix@yahoo•com> Cc: "bitcoin-development@lists•sourceforge•net" <bitcoin-development@lists•sourceforge•net> Sent: Wednesday, May 16, 2012 5:46 PM Subject: Re: [Bitcoin-development] BIP 33 - Stratized Nodes

Thanks for getting this started.

With regards to the specific proposal, I don't believe it's the best option and still plan to eventually implement the original design outlined more than a year ago in this thread:

https://bitcointalk.org/index.php?topic=7972.msg116285#msg116285

Namely that you use a new protocol command to set a Bloom filter on a connection. Only transactions matching that filter will appear in relayed inventory. Blocks that are requested will arrive as a header plus transaction/merkle branch pairs. Clients are expected to maintain and track the block chain as per usual, but instead of downloading the whole chain and then dropping the irrelevant transactions, that filtering is done server side. By strengthening or weakening the Bloom filters you can choose your preferred point on the privacy/bandwidth-usage spectrum. It is a fairly simple change to the Satoshi and BitcoinJ codebases but still allows clients to gain confidence in their balance by examining the chain, and this is true even in the presence of a hijacked internet connection (you can't trust pending transactions that way, but you can still trust confirmed transactions).

The filters would be applied to each data block in each script rather than having a specific knowledge of addresses. In this way you can select for things like multisig outputs or outputs which don't use addresses / pubkeys to authenticate.

I could write a BIP for this alternative protocol if somebody else wants to implement it. I was going to wait until I had time to do both BIP and implementation, but I think some simple optimizations to BitcoinJ can keep its performance good enough for the short term.

> 1) This is cool and useful (but see 3)

> 2) This is significantly less secure than validating an entire blockchain; it's certainly worth working out some use cases here in more detail than just a sample conversation. More on this below
> 3) What about discovery? Will a client now have the chance to look for NODE_STRATIZED clients on IRC? How do you envision a stratized server decides which transactions to relay/store? Or is it just a caching layer in front of a high quality blockchain service? If it is just a caching service, the question of cache hits / misses is an interesting one as well.

Stratized nodes do discovery as normal. Service nodes are explicitly chosen like IRC servers are for IRC clients.

> 4) What are the economic motivations to run a stratized server? Other than cheating people of course.

None. Same as BitTorrent super-nodes, Tor relays or email servers. People don't need economic motivation for everything.

> 5) Seems like a 'send me everything for this source address' is going to save a lot of roundtrip conversations for what I imagine the most common request will be.

That's a bad idea. I prefer to keep each request minimal to prevent resource starvation and simplify the protocol (while shifting the onus onto the client). Also the history can be resolved with multiple services while the data is being downloaded and sorted.

On Wed, May 16, 2012 at 12:46 PM, Mike Hearn <mike@plan99•net> wrote:

> Thanks for getting this started.
>
> With regards to the specific proposal, I don't believe it's the best option
> and still plan to eventually implement the original design outlined more
> than a year ago in this thread:
>
>   https://bitcointalk.org/index.php?topic=7972.msg116285#msg116285
>
> Namely that you use a new protocol command to set a Bloom filter on a
> connection. Only transactions matching that filter will appear in relayed
> inventory. Blocks that are requested will arrive as a header plus
> transaction/merkle branch pairs. Clients are expected to maintain and track
> the block chain as per usual, but instead of downloading the whole chain and
> then dropping the irrelevant transactions, that filtering is done server
> side. By strengthening or weakening the Bloom filters you can choose your
> preferred point on the privacy/bandwidth-usage spectrum. It is a fairly
> simple change to the Satoshi and BitcoinJ codebases but still allows clients
> to gain confidence in their balance by examining the chain, and this is true
> even in the presence of a hijacked internet connection (you can't trust
> pending transactions that way, but you can still trust confirmed
> transactions).

Makes sense.

In an idealized model of a client as a set of private keys, they will want to (a) notice new activity on these keys, (b) notice increased confidence on existing transactions with those keys [confirmations], and (c) be able to submit to the network new transactions. Your proposal covers those bases, I believe.