Bitcoin components.

Lets look at:

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

This is OK as a super high-level overview, but it leaves a great deal to be explained.

First what is the status of this? Is it fully implemented? A working proposal? An approved proposal? etc...

I am working on the peer and the wallet boxes in this diagram. I want to implement my own version because the existing code is completely untrustworthy, it is totally slap-dash and almost completely uncommented.

My questions related to the peer are these:

It is taken for granted that the peer will have a listening component that accepts connections from other peers.

Does it also have an active component? If so, what operations does it perform autonomously?

/me looks at wallet.h and GAH!! Someone reversed the concept of debit and credit! =( (GetDebit(), GetCredit())

Interesting... What is the exact relationship between the wallet and the miner? (The miner is irrelevant to me until I have a wallet I can trust enough to experiment with...)

After reading the dox on libdb, I'm leaning towards an inetd implementation. That is, inetd handles all the dirty work with regards to connections, and spawns off a server for each client. This minimizes the quantity of code at the cost of a wee bit of overhead (maybe 10k per process). Basically, the penguin acts like the KoAC in Katamari Damacy, if you don't do it in the regal old-fashioned way, he rains down thunderbolts on you and you get to write 3 or 4 times as much code.

The upside is that each peer is served by a single completely isolated process with extremely simple and highly auditable code.

A general problem that I haven't seen any documentation on yet is that most of today's operating systems are properly multi-user systems. Naturally, each user would keep his wallet in an encrypted store in his home directory. Are there any standards yet for users to set up their personal wallet daemon that connects to a system-wide peer node and provides an interface for only their own GUI?

On Thursday, July 28, 2011 1:25:46 AM Alan Grimes wrote:

> https://en.bitcoin.it/wiki/Infrastructure
>
> This is OK as a super high-level overview, but it leaves a great deal to
> be explained.
>
> First what is the status of this? Is it fully implemented? A working
> proposal? An approved proposal? etc...

I put the page together a while ago with logical segregation of the current Bitcoin concept. wxBitcoin originally implemented the entire thing as one big blob. Miners got moved out first. Spesmilo took the next step of separating out a GUI. Wallet and node are currently tightly tied together in bitcoind, though it can function as only one of the two just as well. So in practice, it is fully possible to run it like that, but I'm not aware of any de facto use yet.

> It is taken for granted that the peer will have a listening component
> that accepts connections from other peers.
>
> Does it also have an active component? If so, what operations does it
> perform autonomously?

Generally, nodes need to make outbound connections too, at least to get started.

> Interesting... What is the exact relationship between the wallet and the
> miner? (The miner is irrelevant to me until I have a wallet I can trust
> enough to experiment with...)

The miner can really get by with just an address from the wallet-- it needs somewhere to put its generated coins ;)

> A general problem that I haven't seen any documentation on yet is that
> most of today's operating systems are properly multi-user systems.
> Naturally, each user would keep his wallet in an encrypted store in his
> home directory. Are there any standards yet for users to set up their
> personal wallet daemon that connects to a system-wide peer node and
> provides an interface for only their own GUI?

You can run a system-wide bitcoind (or equivalent) to handle p2p connectivity, and have each user run a dedicated wallet bitcoind (with the -connect option) with a unique JSON-RPC port (for Spesmilo) or wx GUI. I don't think there is any automated setup for this kind of arrangement. I personally foresee the p2p being implemented on a SOHO router as the end goal (or even better-- multicast).

On Thu, 2011-07-28 at 01:25 -0400, Alan Grimes wrote:

> I am working on the peer and the wallet boxes in this diagram. I want to
> implement my own version because the existing code is completely
> untrustworthy, it is totally slap-dash and almost completely uncommented.

Yay, yet another person who wants to write their own client from scratch who will get the net code done, work a bit on wallet code and move on to something more exciting. Seriously, at this point it seems everyone and their mother are writing their own, or planning to write their own, client. Bitcoin's code actually can be fairly well segmented into the necessary pieces (CWallet was the first big chunk, more are coming) and comments are being added as a part of that effort. The current code is much, much more trustworthy than anything someone new will start to code, even if they have 20 lines of comments for each line of code. If you want to help the project, please dont write yet another new client library, just look for an existing one you can help with, or help with the original client.