On Mon, Apr 7, 2014 at 10:55 PM, Paul Lyon <pmlyon@hotmail.ca> wrote:
I actually ask for headers from each peer I’m connected to and then dump them into the backend to be sorted out.. is this abusive to the network?

I think downloading from a subset of the peers and switching out any slow ones is a reasonable compromise.

Once you have a chain, you can quickly check that all peers have the same main chain.

Your backend system could have a method that gives you the hash of the last 10 headers on the longest chain it knows about.  You can use the block locator hash system.

This can be used with the getheaders message and if the new peer is on a different chain, then it will just send you the headers starting at the genesis block.

If that happens, you need to download the entire chain from that peer and see if it is better than your current best.


From: Tier Nolan
Sent: Monday, April 07, 2014 6:48 PM
To: bitcoin-development@lists.sourceforge.net


On Mon, Apr 7, 2014 at 8:50 PM, Tamas Blummer <tamas@bitsofproof.com> wrote:
You have to load headers sequantially to be able to connect them and determine the longest chain.

The isn't strictly true.  If you are connected to a some honest nodes, then you could download portions of the chain and then connect the various sub-chains together.

The protocol doesn't support it though.  There is no system to ask for block headers for the main chain block with a given height,

Finding one high bandwidth peer to download the entire header chain sequentially is pretty much forced.  The client can switch if there is a timeout.

Other peers could be used to parallel download the block chain while the main chain is downloading.  Even if the header download stalled, it wouldn't be that big a deal.

> Blocks can be loaded in random order once you have their order given by the headers.
> Computing the UTXO however will force you to at least temporarily store the blocks unless you have plenty of RAM.

You only need to store the UTXO set, rather than the entire block chain.

It is possible to generate the UTXO set without doing any signature verification.

A lightweight node could just verify the UTXO set and then do random signature verifications.

The keeps disk space and CPU reasonably low.  If an illegal transaction is added to be a block, then proof could be provided for the bad transaction.

The only slightly difficult thing is confirming inflation.  That can be checked on a block by block basis when downloading the entire block chain.

> Regards,
> Tamas Blummer
> http://bitsofproof.com

On 07.04.2014, at 21:30, Paul Lyon <pmlyon@hotmail.ca> wrote:

I hope I'm not thread-jacking here, apologies if so, but that's the approach I've taken with the node I'm working on.

Headers can be downloaded and stored in any order, it'll make sense of what the winning chain is. Blocks don't need to be downloaded in any particular order and they don't need to be saved to disk, the UTXO is fully self-contained. That way the concern of storing blocks for seeding (or not) is wholly separated from syncing the UTXO. This allows me to do the initial blockchain sync in ~6 hours when I use my SSD. I only need enough disk space to store the UTXO, and then whatever amount of block data the user would want to store for the health of the network.

This project is a bitcoin learning exercise for me, so I can only hope I don't have any critical design flaws in there. :)

From: tamas@bitsofproof.com
Date: Mon, 7 Apr 2014 21:20:31 +0200
To: gmaxwell@gmail.com
CC: bitcoin-development@lists.sourceforge.net
Subject: Re: [Bitcoin-development] Why are we bleeding nodes?


Once headers are loaded first there is no reason for sequential loading. 

Validation has to be sequantial, but that step can be deferred until the blocks before a point are loaded and continous.


On 07.04.2014, at 21:03, Gregory Maxwell <gmaxwell@gmail.com> wrote:

On Mon, Apr 7, 2014 at 12:00 PM, Tamas Blummer <tamas@bitsofproof.com> wrote:
therefore I guess it is more handy to return some bitmap of pruned/full
blocks than ranges.

A bitmap also means high overhead and— if it's used to advertise
non-contiguous blocks— poor locality, since blocks are fetched
sequentially.



------------------------------------------------------------------------------ Put Bad Developers to Shame Dominate Development with Jenkins Continuous Integration Continuously Automate Build, Test & Deployment Start a new project now. Try Jenkins in the cloud.http://p.sf.net/sfu/13600_Cloudbees

_______________________________________________ Bitcoin-development mailing list Bitcoin-development@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/bitcoin-development


------------------------------------------------------------------------------
Put Bad Developers to Shame
Dominate Development with Jenkins Continuous Integration
Continuously Automate Build, Test & Deployment
Start a new project now. Try Jenkins in the cloud.
http://p.sf.net/sfu/13600_Cloudbees
_______________________________________________
Bitcoin-development mailing list
Bitcoin-development@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bitcoin-development