Much clearer now. Thanx a lot !

@MoonShadow, @Coins!
====================

Unfortunately I know of no specification of the protocol which I understand completely - with the exception of some 10.000 lines of code C++, which I am currently in the process of reading and adding the missing commentaries, but this still takes a bit of time.

The orginal paper of Satoshi leaves many black spots and the Wiki specification describes messages formats but not the overall logic. I still did not succeed in filling in all the black spots.

All in all, your assessment of my situation certainly is correct, which is exactly the reason why I am here in this forum. I want to learn, I want to discuss, I want to improve my understanding of Bitcoin and I want to contribute with what I am able to. I am asking myself what went wrong in our parts of the conversation in this thread, since "zing~" isn't exactly helpful. 

And, yes, I am still having troubles understanding when and how a node decides which transactions are to be packed into the same / next block. Is this done by every individual mining node? If it is, which I assume, I see another issue: Currently less and less nodes are mining on their own and more and more are joining a mining pool, which hands out shares to their workers. Does this mean, the mining pool "supernode" decides on this question? If it does, then aren't we losing the advantage of a decentralized system, since we are having less and less "supernodes" deciding on what all workers are workign on?

I am sorry, if I am wrong with my assumptions. I am trying to understand.

 

Agree and disagree.

We cannot talk about a chance of solving a block without giving a time period during which the machines try to solve the block. 

No machine finds a block within 1 [nanosecond] and every machine finds a block within 1000 [centuries].

But, if we define "chance" as "probability that a machine finds a block within 1 second", of course a beowolf cluster of graphics cards has a larger chance than a 1GHz single core CPU. The "chance" (if defined that way) is linearly proportional to the hashing performance.

That was the part where I disagree with your posting.

The part where I agree with your posting: Solving a block is not a deterministic type of task but a random one. So if many nodes try to solve it and if their random number generators are indepent (and seeded differently), then their "solving power" in terms of "chance" actually adds up. And this is exactly where my thoughts went into the wrong direction. 

Thanx very much for your posting - it made me think and help me understand the problem of my attack.

I hope I got it right now 

@Raistlan:
========

I am sorry, Raistlan, if I was too cryptic.

The idea, as I understood it in that German forum, is as follows:

Assume, Bob001, Bob002, ..., Bob999 all have 1 GHash/sec and Mallory has 10 GHash/sec. Mallory has a faster miner than all other guys on the network but he does not hold the absolute majority (ie. 50% or more) of the networks mining capability.

Assume, bad guy Mallory knows the addresses of the good guys Bob001, Bob002, ..., Bob999.

Moreover, Mallory has 10 BTC. Mallory now designs a transaction of spending these 10 BTC to Bob001 and sends it to Bob001. He designs another transaction of spending these 10 BTC to Bob002 and sends it to Bob002 - and so on. He also sends a transaction of spending these 10 BTC to another Bitcoin address he holds himself.

Eventually Bob001 starts working on sealing this transaction in a block and so do Bob002, ..., Bob999. And so does Mallory, working on sealing the transaction sending the money to his other address.

Mallory will be first in finding a valid block. Not always, but most of the time, since he has the fastest miner.

 

AHA. And I now realize that this probably was the mistake I made in evaluating this attack. Mallory may have a larger chance to find a sealing block than an individual Bob001 by factor 10, but he will not have a larger chance to find a sealing block than ANY Bob*. Of course he is competing against all other miners combined. Thus, chances are much higher that there is just ANY Bob* to get his block done first and so Mallory's plan will not work.

I hope this is fine by now ?!  


@Maged:
=======

Thank you very much for pointing out the Finney Attack: I was not actually suggesting to accept a transaction without waiting for confirmation. I assumed that Mallory would manage to produce a longer block chain with his version of the transaction faster than all the others - but this is plainly not true, there was a reasoning mistake by my part here.  

Sorry. I just corrected two typos which might have been misleading. What else is not clear? Or am I too short?

The idea is that Mallory is doublespending on a massive scale - sending a different transaction to EVERY participant in the net. He will win the longest block, assuming he has the largest mining rig.

Hi there,

in the fallout of the recent theft of 0.5 M there was an interesting discussion on a german board and a suggestion for an attack. I am not sure if we can withstand this attack and thus post it for comments here.

Mallory has a good mining rig / pool. He has by no means the majority of the hashing performance but he knows he has the largest or second largest hasing performance.

He now sends to every participant a different tx and starts building a block on yet another tx. Everyone will solve HIS version of the tx and Mallory will win, assuming nobody has more hashes / sec than him.

Note: We do not assume absolute majority of hashes, but (only) being the largest participant.

According to my understanding the attack would work, although it might not be practical.

Opinions?

Ah, ok. very helpful !

So it is essentially a more compact form of storing large values for saving space (and probably my attempts to bite the bullet and understand the peculiar details of it are a bit futile).

Hi there,

I am having a hard time understanding the CBigNum::SetCompact and GetCompact functions.

Ok, this is about some compact representation and I can see it in the code how it is done. But I am lacking theintuition behind this.

Why is such a representation necessary at all?

And why in exactly *this* manner?

Would be happy for some hint here.

https://en.bitcoin.it/wiki/Script reads:

The script for a typical Bitcoin transfer to destination Bitcoin address D simply encumbers future spending of the bitcoins with two things: the spender must provide a public key that, when hashed, yields destination address D embedded in the script, and a signature to show evidence of the private key corresponding to the public key just provided.

and I am unfortunately too stupid to understand this.

The spender wants to transfer money to destinatin address D. This destination address is the hash of the public key of the account to which the money is transfered. Right ?

But how can the spender produce a signature to show evidence of the private key? The spender does not have the private key to the "public key just provided", since this is the place he sends the money to.

So, obviously my understanding is wrong here.

Where is my mistake ?

Agreed ! 

In a forum on manipulation resistant, censorship immune digital currency - we really need moderation, manipulation and censorship, don't we? 

The original poster got me mildly amused. That's how markets work, by pumping up panic and observing the reaction of the crowd. We should not fear kids with bit coin panic, since it is just their action which keeps the currency stable in the end. They are not smart enough to really manipulate the course, and they cannot cough up sufficient trading volume as well. But they get people thiunking about the course and the whole concept - and this is very good.

I am using the catalyst drivers from the AMD website. Are you using the apt version?



to me it looks like the second device does not do any calculation at all. At least: The "verification error message" is triggered when a rough plausibility check of the results shows some bits to be zero which should not be zero. It looks like the driver ignores the params to the GPU and just returns default or zero results even without proper error message.

However: There are miners using more than 2 odr even 3 6990 GPUs so I assume there is a trick. 

Hi there,

today my new 6990 LCS from Powercolor with water cooling came.  

While I was unpacking, a phone call came in by the shop where I had ordered yet another watercoold 6990 LCS from Powercolor. The guy told me that he can't deliver since they ran out of stock and - what's more important - they probably would not get new ones, having way too many burnt cards and warranty work is getting a problem for them. They could, of course, send me a do it yourself watercooler kit, but then without warranty for the card, and actually, they would recommend against it and suggest the air cooled one.

So I told him to put the order on hold and I would be back to him as soon as I had some more experience with the just-unpacked card.

There I went, putting together the new cooling circuit, leakage test, filling, test run and all that stuff. When I finally booted with the 2 cards (one water cooled 6990 and one air cooled 6990), Win 7 started to do some reconfiguration and driver installation on its own when all of a sudden there was this funny sound like spraying water, followed by buzzing sound. Screen went blank and air went smelly.  

...and there it went the 6990 LCS from Powercolor.

The card is still completely dry, not a single drop of coolant. There was no leakage or stuff, but the sound and the smell spoke a clear language, as did a slightly brownish color.

Now I understood: What they guy on the phone meant was not an overclocking fatality but a burn-for-no-obvious-reason fatality. I did not even get so far as to have the drivers installed completely. Only thing which caught my attention and surprise was that device manager only showed 3 GPUs instead of 4 = 2x2.

Any similar experiences regarding Powercolor ?

It took me a while of configuring to get the air cooled 6990 workign again. Initially control center reported 0 clock frequency and the card did not show up properly.  





 

Agree.

But the faith should not be dependent on many people MINING Bitcoin, but on many people USING Bitcoin.

I can relate to your feelings, since I discovered the hype for myself a few days ago. I realize that by the time my next deployment step with these shiny new 6990 watercooled miners is running, I will probably be heating my appartment with it and not making profit at current rates.

The real faith comes with people USING it to buy and sell goods and this needs a totally different form of software development.

Can't we always produce more coins just by splitting them further?

So maybe we should look into splitting techniques and into wallet improvement.

On the other hand: A thing which scales exponentially...always is very bad to adapt. So let's see.

And: If we split coins and prices go up...the early adopters in mining will make a fortune...and this might lead to greed and bad temper in the community. Still, it's granted. The guys took the risc !

I just started mining these days and am a n00b - but I had an interesting phenomenon on my first day you might consider useful.

For some stupid reason i let a piece of plastic sit on the top of my 6990 - and some minutes later rectangular shaped parts of the screen went black. Temp was at 85 Celsius. I turned off, let cool down and restarted. The areas where back again.

I assume it was a partial memory failure... :-(

 

Being a bit deperate at this end...having worked 2 days on my brand new 6990 and unable to use the second GPU.

This is an Athlon x2 on an ASROCK 890 GX Extreme 4 with a Sapphir 6990.

OS is latest Debian 64 bit via netinstall, ATI Driver Package is 11.5, Catalyst is 10.9 and SDK is 2.4

What I get after the usual fighting against docs and google pages is a nice 330 MH/s on a poclbm on device 0, but on device 1 is get the verification error message.

I did check the usual stuff, like DISPLAY to :0 and turning off crossfire and all kinds of things, but it still is the same. I read that most people here are using SDK 2.1 so I tried to downgrade to 2.1 but here I opencl does not see the GPU at all.

Any ideas please?

Or, any confirmation that someone is mining under 6990 and Linux successfully - and could you send some params such as SDK versions, driver versions and stuff you had to do...

Since I am looking forward to upgrade this to a 3x 6990, Windows is not really an option due to the 4 GPU limitation.

Best,
Forp.