Bitcoin Forum

This thread is a spin-off from:
https://bitcointalk.org/index.php?topic=99497.msg1154935#msg1154935

Please use it to discuss the implications of the emerging field of ASIC mining and the role of ASIC hardware companies. Mentioned topics are interest of conflict between using the chips and selling them, secret algorithms and back-doors.

Enjoy!

I don't understand all the fuss about "secret" backup hashing algorithms (in the ASICMINER thread (https://bitcointalk.org/index.php?topic=99497.msg1154765#msg1154765)), and I don't see a problem with implementing a backup algorithm in hard-wired miners (ASICs). For years we've been mining with configurable hardware (CPUs, GPUs, FPGAs) that has practically infinite number of secret algorithms "embedded" - simply reprogramming the devices to a hashing algo of your choice would do. Why is the thought of hard-wired backup hashing algorithm in ASICs so scary?

The thinking goes that if the algorythm needed to be changed to one that ASICs cant use and they are the majority of the network then they could vote not to change the algo simply be owning the majority of the hashing power.
We would get stuck on a broken/substandard algo because of it.

What does "who people want to pay" mean. Can you define or explain this statement specifically?

As far as I understand it this is correct. But turning the mining community against each other would probably result in a hard fork, where the economy gets splits into two coexisting realities: In one chain my coins might be spent, in the other one they are not.

Why would you exclude a broken SHA256 scenario? It's a perfectly valid reason to have a backup hashing algo in case the first one breaks.

For example, if you want to pay BitVOIP for some nice VoIP services*, the only thing that matters for that transaction is what Bitcoin protocol they are willing to accept. Inevitably, for Bitcoin to work at any scale, it is the merchants people want to do business with the most who matter.

* No, I don't know anything about BitVOIP or anything like that. I just quickly peeked at the Trade wiki page for a quick example I'm not biased on.

I'm assuming that if SHA256 gets broken, any backup variation of it automatically is also broken. The algorithm might need to change anyway, but it would break all ASICs.

Then it should not be a variation, but a completely different hashing function.

No. it requires both.

Yes if no one uses bitcoin, then there will be little future for it.

However, if the blockchain doesn't get securely verified - no one can use it :P

As I said :)

That probably doesn't come free. Since it is impossible to know in advance just how SHA256 will be broken (if it ever is), it is also probably not worth any cost to try to add a complete alternative to it, since it could just as well also be vulnerable.

Ok, I think I get it now. I just wasted my time with this whole discussion. Everyone pretty much agrees, even Luke-Jr - but he can't stop there but goes on meaningless and confused tangents. I'm out of here.

LOL - do some reading about hashing algorithms please :)

In the other thread Luke-Jr twice ignored my request to show where such an advisory was given.

I have yet to see that this entire fracas is anything other than his fevered imagining.

<sigh> What fact?

I see no fact in evidence.

Merely a fanciful statement you have made and are apparently unable to support via any citation.

Exactly. Changing the algorithm would only benefit the few big money players, who can build a new ASIC chip fast, and hijack the market. That would be the single most stupidest thing that Bitcoin can do it for it self, destroy all the existing ASIC diversity.

ASIC is the end-of-the-line, and we need as much players to that field as one can get in order to secure Bitcoin's future. There is even a OpenAsic project, why people who are concerned of concentrating of power doesn't support that? Give some BTC love for it, and all is solved? But noo, they will shoot them selves in the leg, because they hate that their GPUs are becoming obsolete. Or they are so stupid that they would agree to hand Bitcoin over on a silver platter to a few rich players.

Support OpenAsic, or start your own (to only mine with them, for all I care). We need all the hashrate we can get, in many hands that's possible.

Luke just doesnt want to move to litecoin with his GPU's when they are made redundant  :D

Well, as I have said above (and elsewhere) there is reason why the algorithm could need changing.
Firstly, if the sha256 was broken.
i.e. if someone works out how to factor the hashing process enough to solve blocks rather than hashing (currently) on average 1.05x10^16 double sha256's to find a block.
If that happens, then of course sha256 much be changed - of that there is no doubt.

Though, I have already give a reason why the current hashing process needs to change (not the sha256 algorithm) here:
https://bitcointalk.org/index.php?topic=89278.0
However, I've no idea if that would affect the ASIC implementations, coz it would depend on if they have optimised the double sha256 (that give an extra 6.25% performance) as has been done with GPU hashing, or not.

As soon as BFL ships the ASICs, they have no control over them. Their own customers will be securing the network against such an attack. If they tried to pull off such an attack before shipping, the Bitcoin community could just switch to an algorithm their chips don't support.

As long as that is true, it's obviously not going to happen. The risk of it happening only comes with someone having ASICs online before anyone else.

Quit putting words in my mouth, kthx.

... and how long would that take?
The last soft fork took forever and failed dismally - massive orphans (and chains of orphans up to 4 long) for almost a week.

If someone had the hardware to do an attack they could do it and make a profit before anything would change in the bitcoin code ...

Which is also why I ALREADY said that this required foresight and planning ... but of course that planning WONT happen.

They have no control over the ones they've shipped, true. However so long as the amount of ASICs they've kept back is a reasonable amount larger than the amount they've shipped they'll still be able to carry out the attack, and for various economic reasons we might expect this to actually be the case for any manufacturer of Bitcoin mining ASICs.

Would it not make best sense to push back delivery as much as possible (talking about any ASIC rig producer), for as long as possible, until reward halving? Burn in testing is more profitable than delivery into a small market. This had to be a balanced act as not to have customers cancel their orders or ruin cred.

I suspect the latest steady rise in hashing capacity are from ASICS from multiple sources also the less public ones.  No one would buy new gfx cards now? Second hand gfx market would have zero net effect on global hashing if they are bought from other miners. Investing in an ASIC rig, and mining at a higher difficulty, will make your coins more dear. An outside negative influence on the economy, or the technical side of having a non upgradable rig, could start a sell off.  If the market gets really flattend by regulation or otherwise, running an ASIC could be very profitable, if like 70-90% of miners goes of-line in fear of criminal charges.

There is an obvious reason Butterfly Labs does not want to mine themselves: In the long run, mining is not very profitable. It is expected to converge on the cost of bandwidth/electricity. There are only so many bitcoins to go around; only so many block processed per day. In the long run, Butter fly labs can make more money selling chips to miners trying to outdo each other.

Please don't confuse long term strategy with short term strategy. On short term it is very rational for BFL to do a bit of self-mining. And there are different ways to market that: Quality control, Shareholder profit (ASICminer), Burn-In, .... 
Why would BFL want to let a highly profitable business opportunity pass? All they need to do is throttle the hardware selling a little bit and keep each chip at BFL for quality control and burn-in for e.g. 2 weeks. Do that round-robin and you got yourself a free mining facility.
Of course you're right - if the market saturates the profit from burn-in and quality control is negligible. So it doesn't make sense to build a huge mining facility.

It's a food chain and the chip producers are at the front of the line. Mining facilities are second in line. People who hold mining bonds or contracts are third.

If they really want to do a burn-in, they can use test-net.

And let an opportunity to make profit pass? How is that rational?

Why should they do that? It doesn't make sense at all since its about business and not fairness or whatever.

Merely due to the company's honor?

I do seem to recall they clearly stated they would not mine with their devices prior to shipping.

After all, they have a reputation to maintain.

-- edit --
Nope, I was wrong.

It'll be interesting how that "promise" plays out. I think the damage to the reputation wouldn't be huge. Also there's plausible deniability: "don't have enough manpower to ship above a certain rate"... and in the meantime, why should the devices stay idle?

I expect some sort of excuses. But that's fine. I actually think it's fair for BFL to do self-mining. After all it's the privilege of the innovator in a free market to keep the profit (and pay taxes...).

The only thing I think is not fair is their pre-order strategy: They likely used the pre-order money to finance the chip production, which is very much in the gray area, because the customers didn't agree to a risk of default, should the chip production fail.

Citation needed. I don't know why they wouldn't, the chips certainly need burn-in to eliminate the bad ones.

If they start Mining with one of their $30,000 (1Thash/s) mining rigs, it should show up on Bitcoin Charts (http://bitcoincharts.com/) (as 5% of the network).

No, chip testing is quite a problem in semiconductior industry since  the are only a few seconds per die to decide whether it meets the specs or not. So complicated devices have build in BIST (build in self test) and even RAM chips have such provisions. Since SHA is quite regular and each flipflop is toggling every second  clock cycle it would be take only very few hashes where you compare the output  of some known input patterns against the precomputed output pattern.

Of course  there are other points of failure bur this would require them their other equipment, for instance their power cables at maximum current in a climate chamber.