New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Actually I think he is just sharing things that he has read from this top cryptologist guy and you are trying to attack his ideas which aren't his to begin with.
@Corey
Based on what I see based on his own web site his job and expertise seem to be in the area of system administration and web development. While I don't feel a need for advice in system admin. my knowledge in web development is quite limited and I will gladly and with interest listen to what he has to tell us about web development should that come up.
You'd be surprised...Linux, all the BSDs, Windows Server, etc. are all POSIX-compliant. But POSIX is a pretty low bar.
Warehouse management is part of my day job and saying "not used at all" is not quite accurate. The thing with warehouses is that they're passed around like cheerleaders on prom night. Even the big supply chain companies typically rent warehouses on a term contract basis.
Warehouse owners try to have the buildings wired up for what people typically need to make them generically attractive. So your typical warehouse may have all kinds of industrial hookups (particularly if there have been multiple previous tenants), but only part of them may be in use. Of course, some warehouses have special needs - e.g., refrigeration, etc. But even dry goods require power for dozens of wireless access points, network gear, opening and closing those giant dock doors, etc.
You could certainly rent a warehouse and only turn on the lights. Some people rent warehouses primarily for the space rather than the warehousing facilities (e.g., artists, movies, people storing/organizing for events, etc.) But most warehouses have the infrastructure in place to consume much more power.
Windows can fork?
It doesn't have to in order to meet POSIX 1.
https://en.wikipedia.org/wiki/Microsoft_POSIX_subsystem
As Wikipedia explains, in order to get some government certifications or approvals, Microsoft had to have a POSIX-certified OS and so they did.
The entire POSIX effort was always a deluded mess. Every time the Unix world gets together to standardize Unix it never works (Anyone still using OSF/1? How's that LSB these days? Any updates since 2011?)
There's this mindset that if we can just get everyone to agree on one set of APIs and command flags, anything can be ported anywhere and Unix will take over the world. Instead, what happens is that committees argue until the final product is irrelevant.
What actually happens is that weaker Unices got trampled in the marketplace and vendors certify against specific operating systems instead of standards. I've never in 25+ years seen a product that is "certified for POSIX" - they're all certified for Solaris 8 or Red Hat Linux 6 or whatever.
Cryptocurrencies making ASIC mining possible are kind of idiotic. Just wanted that out of the way.
Another high quality thougt, how is something that’s controlled by a few large entities with huge farms more decentralized than the fiat currency controlled by politicians we vote for?
I'm sure that the only known sentient species in the entire universe hasn't spent the last few years furiously burning up resources because individuals want to get rich, as opposed to using this power for something like curing cancer. I'm sure that wouldn't happen.
"POSIX"...
Confirmed. For the system info part of vps bench I had lots of "fun" using wildly different sysconf & friends Posix "standard" calls. linux is a particularly ignorant party and OpenBSD takes a lot of liberty too (but also offers full and attractive functionality).
Uhm, just to be fair: EVERY algorithm can be put into an ASIC.
Absolutely.
NO! NOHO! How could something like that even cross your mind?
Btw. how sure are you about the "sentient" in "sentient species"? Or are you talking about aliens?
For one, I can not really speculate, because deep down I am heavily influenced by my own selfish wishes of playing games with high graphics. Therefore I must allow myself to think (or hope) prices will drop in an attempt to sell it all on online auctions and get as much as possible, once cryptomining (with GPU) is over.
Well I said known. I was excluding
Interesting, i knew about MS POSIX subsystem but i've just always thought it was kind of a drunken idea by MS to get Unix software to run on their sorry servers which obviously didn't work all that great missing fork since imitating it's behavior using threads is a bloody mess. Didn't know this was actaully enough to become compliant. No surprise then that expecting anything but the mere basics from a POSIX system is such a lost cause.
Iirc there is actually a crypto currency that uses the computation power of their miners for some kind of cancer research. I didn't feel like checking their claims though so i have my doubts as to what extend this is simple marketing as opposed to anything usable.
In theory, sure. Practically at least high RAM requirements seem to be costly and therefore not economical in relation to mining and at least some currencies try to use this to cripple ASICs. Supposedly FPGAs are way better at dealing with high RAM requirements and still beat GFX cards in terms of efficiency though.
Also i don't see how it really matters what actually fills the warehouses. Be it ASICs, FPGAs or GFX cards. For the average guy hoping to stay competitive using his gaming card on the side it'll always be vastly destructive. And going further crippling GFX cards too doesn't seem to be a solution either since as i mentioned above those systems sooner or later get overrun by botnets trying to cash in on their zombies.
And finally it seems all those tries to deter technology X don't seem to work all that great. Supposed CPU only coins are found to be GPU minable all the time and ASICs get released for coins that should be ASIC resistant according to their authors.
Just look at Monero. They've just switched their algo to at least break ASICs for now and other coins with similar claims are pretty much switching on a regular basis. A pretty bad thing to do for something that is supposedly decentralized.
No, I'm not into aliens and UFOs. So you insist on KNOWN species. OK, dolphins then maybe?
I guess you confuse things. It's the GPUs that have problems with lots of memory due to their high core count. 8 GB for example sounds like plenty but when those 8 GB is for 2000 cores/units it's not that much and an algorithm that uses 16 MB can't be (fully) parallelized (and often not at all due to diverse factors). Modern password hashing and key derivation functions typically make use of that fact. Both FPGAs and ASICs do not have that problem unless they follow a high parallelization route.
Yes, right and one major reason for me doubting the "democratization of currency" of crypto currencies.
One can't really "break" any technology with crypto (without getting completely unusable for the normal user). It is THAT problem zone where the war is fought.
Whatever crypto comes up with FPGAs and ASICs can, at least theoretically, follow. But luckily for crypto designing and building hardware virtually always takes much more resources, money, and time than changing software.
But there is another decisive issue: WHAT is the driving interest? If it's money/profit the game is very different from when it's eavesdropping and friends by intelligence agencies.
If it's money/profit our job (in crypto) is basically filtering and equalizing; filtering as in "making 'special golden routes' e.g. for resourceful miners more difficult and expensive" while at the same time (at least ideally) keeping the game equal enough for the vast majority of normal players (the equalizing part). It's usually about drastically narrowing the gap between some normal user vs. a resourceful user with say a GPU farm. Formerly that gap could be 10 to the 3 and even (much) higher. With a good and well parameterized PHF like say Argon2 there still is some gap but it's much narrower and most of the top is cut off. So, the normal guy with a modern notebook might need 0.5 seconds but even the high end guy (say a resourceful criminal gang) can't do it 100 or more times faster. In my own projects typical gaps are 1 : 10 - 50.
And as it's about money/profit the other side is seriously limited too. At the same time though that's one of the points where it's NOT democratizing: if you are a very resourceful player you CAN escape the gap and get an advantage of 1 : 1000 and more.
But then security in general is never really about the "impenetrable vault" but about driving up the cost for the intruders, to win time e.g. to bring in other types or simply more defenses.
If the driving interest is what typically drives intelligence agencies the game is very different and one very major reason for that is that the cost factor is of little concern. There are other relevant factors too but this post is already long so I keep it to one particularly unpleasant one: The very same agencies are VERY DEEPLY connected to the white hats too via research grants etc. In other words, if say an NSA does it's job well and intelligently then it's very hard for us white hats to not be behind them and maybe even ahead.
That sounds logical and i guess you are right. Quite a couple of crypto currencies that try to penalize ASICs try the same for GPUs so i might have very well mixed up something. I have read about the actual RAM chips for the ASICs being expensive though but then this really isn't my field and i am pretty much just an interested bystander when it comes to stuff like building ASICs or programming FPGAs.
I guess there is a bit of a misunderstanding here. What i ment by breaking is merely rendering currently existing ASICs useless. It might have a bit of a broader effect by making the money spent on developing those become more of a gamble but if this is enough to produce an unfavorable the risk/reward factor is to be seen. Imo it's probably just going to make ASIC makers become more secretive and not sell to the public at all. Also i have a feeling that the algo wasn't changed as fundamentally as to not allowing at least parts of the existing design to be reused.
Exactly.
Interesting. So there is actually at least some scientific validity to the claims made by Argon based coins. Not sure about which version though and judging the used parameters is beyond me. I am curious what you would say about the various Yescrypts. Are they even a thing in a professional environment?
Yes, I have. It has significant technical limitations, and is not a full replacement for on-chain transactions.
Indeed. Anyone who knows how lightning actually works will know why it doesn't work
It's the algorithms and implementations (on which the crypto currencies and other applications are based) that try to "filter and equalize". To do that one must always take into account what kind of devices opponents might use; typically nowadays the answer is (common) CPUs, GPUs, FPGAs, ASICs, and distributed resources (e.g. botnets). CPUs are relatively easy to defend against but important because everyone and his dog have one or more available but also because they are (usually in a weak form like e.g. home routers, IOT) used in botnets. GPUs are relatively similar but with a large amount of units. FPGAs and ASICs are the hard target because (a) you don't know little about them (unlike a CPU/GPU no fixed functionality and parameters (e.g. registers)) and (b) especially FPGAs are flexible and can quickly change and adapt.
There are ways to at least somewhat mitigate but that would get too complex here.
ASICs are not the nightmare. FPGAs are, at least for most crypto algorithms because they can adapt quickly. That said, we are beginning to have algorithms that are "chameleons" and change steadily which thwarts most FPGA and ASIC attempts.
Yes and no. In theory yes but in real life mostly no. Reason: Especially for a crypto currency (but also for many other applications. Just think "TLS") one simply can't change the algos at will and frequently. Plus ASICs are typically not made early on, say when a new crypto currency comes up, but only after a while both because the process itself takes quite some time and because making ASICs is only worth it for a major and well established crypto currency.
Depends. Argon(2) is an excellent PHF and KDF with good m and t. But like every crypto algorithm it must be very well and fully understood, properly implemented and used. So "it's Argon based!" is more of a hint than a guarantee. One can build shitty mechanisms with Argon too as well as surprisingly good ones with boring old SHA-2.
Re. this or that crypto currency I will not or very rarely make statements. The ones I did study closer either because I was asked to or because I myself wanted to know were ALL between questionable and rotten and I have more interesting things to do than to study/analyze each weeks new super crypto currency. Let me put it like this: Based on what I know I'd say that there are WAY LESS capable people who actually could do the development and design incl. protocol properly than there are crypto currencies...
Plus to the best and most up to date of my knowledge there is no crypto currency that can actually make a sound and solid statement about their safety and security. "It's based on Argon2!" for example is a nice hint and Argon2 is a nice algo but the usual laws of the field are true and valid for crypto currencies too. To say "X is safe and secure" one must be able to prove that, particularly wrt to the implementation of the whole software, server and client side and gadgets like Android interfaces.
Yes, Argon is secure - but so is even RSA beyond even just 1k size, yet quite some RSA based software has been broken/hacked because Argon or RSA or whatever no matter how good they are are but one (1) component of a code base that is much larger. OpenSSL and lots of software based on it should have tought us that lesson brutally enough.
So when I hear "X is safe and secure" my first reaction is always "Show me the proof" which virtually always leads to very quiet people after some "It's Argon2/ECC/AES... based!" blurb.
Based on my concrete experience I wouldn't touch a crypto currency with a pole.
Agreed. As you say yourself below the ability to modify algos is not all that useful to crypto currencies though. Not only are the transitions generally messy but they also need a centralized authority to coordinate the transition which goes against one of the fundamental design goals of most crypto currencies.
Also fully agree. Monero was obviously only targeted because it was profitable to do so. Also what i said above applies. I hope you don't get me wrong. I am not seeing algo changes as any kind of sustainable solution. It's a whack-a-mole game with no end in sight.
Of course. I didn't want to make it sound like i was advocating those currencies on the simple fact that they are using algo X. I was merely acknowledging that if one was to put their choice in a vacuum and assumed a reasonable implementation there was at least a chance for their claims to hold water. That's also why i brought up Yescrypt whose users are making similar claims. I was simply curious if it was a name that would even be recognized by a professional.
Impossible to argue against this. There are 1000s of them and a good portion is likely to be developed by kids.
I hope you don't assume i am that naive. Beyond that, sure, people take a lot of the marketing nonsense for facts when they should at least be skeptical because they lack the ability to verify any of it.
Don't worry i think this is already crystal clear to anyone following this conversation. I don't use any either. Well, i might have 1 or 2 shitcoins from playing around with it and wouldn't be opposed to using crypto for small transactions if it were convenient enough (it's really not - well, at least where i am right now Bitcoins are sold at gas stations but afaik the conversion rate is pretty bad). Would i risk any serious money on it? Certainly not.
Potential misunderstanding of [any application] changing/modifying algos vs. algos that are "chameleon".
In theory yes. Practically though the problem is rarely this or that algo but rather bad design (protocol!) and implementation and operations. But you are right insofar as quite many particularly poor crypto currencies OBVIOUSLY made questionable and uninformed algo choices. Having well cryptanalyzed solid and well chosen for the task at hand algos is a good start at least and Argon2 often meets those criteria.
I try to assume very little. In fact and as this discussion shows I did and do not even assume a lot when you were VERY mistrusting and unfriendly when I came here.
So, no, I do not assume that you are naive.
I think of humans as the only sentient species on Earth. But I profess genuine ignorance on the subject, and also disinterest in the debate.
Guess so if you say there is a difference.
Well, i don't think i was particularly mistrusting (i still think the bit about MS was kind crystal ballish future telling) but unfriendly pretty much fits. I'd say asshole mode but oh well... It's partly this place partly my nature. Also i have a lot of time to pass right now so excuse if i troll a bit here or there. Hope you won't hold a grudge against me.
Explanation: Some say crypto currency changing from "1000 times SHA256" to Argon2 is one thing.
A crypto algorithm that is "chameleon" however is another and quite different thing.
Sub-explanation: Most people think an ASIC is something like a somewhat specialized CPU when in fact it's more like a hardwired FPGA. So unlike a CPU ASICs don't like to (and usually can't) run changing algos or do so only at a severe penalty.
Don't worry, just a little friendly trolling. I do not hold a grudge against you as this discussion should quite clearly show but I was really quite shocked when I came here.
Btw I'm not in the mood to be grumpy anyway because I just cut down the time for some (good quality) random function by almost 30% by hand unrolling some loops and using some (clean) tricks with ifs at an unfortunate (read costly) place.
GPUs and FPGAs devote a tremendous amount of silicon to parallel multiplier hardware (floating point even, in the gpu case), which afaict is not used much in crypto mining. ASIC miners eliminate that. If someone made a gpu-like board without the multipliers it woud be several times faster at mining for the same die area as as a gpu, and in that way get miners out of the market for normal gpus.
Can you elaborate? It's already been launched on mainnet.
About the only people making good money on the crypto craze are the people providing services and making hardware for it. It's just like any other gold rush in that respect. Relatively few make much money doing the actual mining.
It's quite complicated. But essentially in order for lightning to work, you need to open channels, and commit funds to the channel.
It doesnt make sense to open chanels direct between buyer and seller since that doesn't really solve anything as you need to do that on the blockchain any way. In order to acheive massive scale, you need semi centralization. You open a channel with a hub of sorts, and they have many channels open too. That way transactions can be done off chain and cleared to the block chain only when you need to.
There are many problems with this, but one of the biggest problems I see is that funds need to be committed to the channels. So if you want to process $1M worth of BTC, you need to actually have $1M committed to your channels. You cant just move fund around between channels either. The maximum transaction size is the maximum amount committed with each channel opened.
So I'm a hosting provider and I want to open a channel to RECEIVE funds. This channel operator needs to commit funds to the channel in order for me to actually receive anything. So maybe I tell him I'll do $10K /month in transactions and we can clear to the blockchain monthly. So the operator needs to commit $10k to the channel. An enormous amount of capital needs to be locked up to keep these channels open.
There are a bunch of problems but I can't be bothered to write it all out.
Lightning network can work only if a lot of hoops are jumped through first. And I don't see the point.
Yes depending on what you mine most of this is unused (ETH eg. uses memory, equihash GPU), however what you describe is more of a "limited FPGA" or VEEEEERY simplified ASIC design.
The gain in ASIC is power saving, dropping all shaders from a 1080 and removing the memory interface will not get it any considerable power savings at the same equihash rates - ASIC design is from 0 or from an existing ASIC base of yours toward the calculation goal, not backwards from eg. a GPU. This is also why the ETH miner is crap, they just failed on the watt:hash goal unlike before, not entirely and it is usable, but by % it is WAY worse than what Bitmain had before. Then as next step you can talk with your FAB about scaling it further down to eg. 18nm-.
The extreme cost is R&D, the silicone and FAB time saved is... not too relevant.
Keep lastly in mind ASICs are not 12/14/16nm but mostly 20nm or more; yes this is cheaper at the FAB and with the increase of performance over a GPU (plus smaller size of DIE) it makes sense but is not the best possible, it is the best cost:value:tdp cross.
(Disclosure: i own at shares in at least one GP ASIC manufacturer (it is not Bitmain) and multiple FPGA ones, including ones now owned by Intel that converted to Intel shares))
You open one channel and you can pay anyone on the lightning network with it not just the one person you are dealing with now. Calling it blocked funds is a misconception. All your dealings will be via lightning so how is it blocked funds? You can pay whomever you like with that one channel and anyone can pay you too.
Also combining funds in different channels (if for some reason you opened more than one) is something they are working on. It's going to happen.
You don't need to spend money to receive money. That's a misconception. You only need to share a public key which costs you nothing. Furthermore you will almost never close a channel since like I said you use one channel for everything. You only resort to settling on-chain in the event of a dispute.
The point is to scale bitcoin beyond what is possible with on-chain transactions. Instead of committing every transaction to the blockchain you only commit some of them. The rest are signed multisig transactions that are exchanged between the participants off-chain. The blockchain only gets used to anchor your payment channel at the start and as a arbitration mechanism in the event of a dispute.
the shares of nvidia and AMD fall because many sell old cards.
What? No no no.
The only way this scales is if there is like 1 hub and everyone opens a channel with that 1 hub. So in theory, you can transfer funds to anyone through 1 channel if everyone is connected to that 1 hub.
If you have multiple hubs, then hubs need to be connected to each other.
The amount of funds you are able to transfer from A -> B cannot exceed the amount committed to the channels.
Assuming a simple scenario, where there is just 1 hub, and 2 end users (A and
, the only way user A can send user B 1BTC is if the hub has committed 1 BTC or more to the channel with user B.
Maybe A is a buyer, and does not expect to need to receive any funds. In order to spend BTC on lightning, he needs to commit BTC to the channel, but the hub does not. So if he puts in 1 BTC, then that is the maximum he can spend.
User B is a seller and expects only ever to receive funds. So they as a seller do not commit any BTC to open a channel. But in order for the hub to actually transfer funds to user B, the hub needs to commit some BTC, which cannot be accessed or used for anything else while the channel is open. In order for user A to send 1BTC to user B, the funds committed by the hub to the channel with user B must be 1BTC or greater.
So lets look at this at scale.
If a hub wants to process transactions for sellers, they are limited to the number of sellers they can support based on how much BTC they allocate to each channel and how much BTC they have. And so long as those channels are open, you dont have access to the funds, which has cost.
If on average the channels have 1BTC committed to them, and you have 1,000 'sellers', that's 1,000 BTC, which at the moment is worth $6.5M. That's $6.5M locked away and inaccessible, which actually is a cost of sorts.
I did not say spend. I said commit. And someone needs to commit some funds. As mentioned above.
It doesn't really scale that well though. You can't spend more than what's committed to a channel so although you can push quite a few transactions off the chain, you still HAVE to go back to the blockchain once in a while to increase channel limits or clear out funds.