5 ms·

Their argument for using ternary is that it is 8 times more efficient than binary and it is very important for iot devices which need to be have low energy cons

Their argument for using ternary is that it is 8 times more efficient than binary and it is very important for iot devices which need to be have low energy consumption and a long life. Whereas the conversation from ternary to binary would add only a 10 % additional usage. They are trying to push the ternary to see if adoption can take off. All it needs is a few thousand logic gates on any traditional chip.if it doesn’t get adopted,they said they can always go back to binary.

imglorp 9y ago
> 8 times more efficient than binary They're claiming this, for what operation? And similar question for the 10% usage claim. I suspect storage will be the same when you get done packing into machine words and storing those, yes? Unless they're not using all the bits in a word... Any independent references for ternary efficiencies for any operation?
- smrtfckr 9y ago
  IOTA is built for the internet of things and more specifically for sensors operating in the field. Perhaps with limited acces, perhaps with long intervals between service periods. Generally, you would't care about an 8 fold efficiency increase when you're hooked up onto a powerline but remote sensing equipment does.
  - imglorp 9y ago
    Great, someone else quoting 8x efficiency. Maybe you know about the claim? We understand the motivation, which keeps getting mentioned, but what I think we all want to know is: 8x efficiency _of_what_compared_to_what_?
  - PinguTS 9y ago
    Such devices are using ARM Cortex M0, M1, or M3 processors. Maybe even lower power Atmel stuff like in the Arduino. Such devices are connected via either expansive uplinks or slow uplinks like with LPWAN. For all these applications IOTA is not feasible. It is not feasible because the ternary logic needs to be emulated on a processor that itself has not even enough beef to do cryptographic functions by itself. If they have cryptographic support like the e.g. the SE series from STM with a Cortex M0 as used in the Ledger Nano S, then this is optimized for standard cryptographic algorithms and not for some newly invented stuff. Also if we talk about IOT sensor networking with LPWAN (LoRa, SigFox, ...) than bandwith is a real issue. That means collecting transactions from others and doing PoW on that before sending of your own transaction is completely unrealistic. The remote sensor just wants to transmit its data. But that would mean the LPWAN Gateway needs to be a full node doing the PoW. But LPWAN works, that everybody can listen to transaction request and so we are back to the collisions problem.
- etunity 9y ago
  I am not sure about 8 times efficiency and 10% argument. The founder said that in his interview yesterday. https://m.youtube.com/watch?v=GwhJQ67zxbg https://m.youtube.com/watch?v=GwhJQ67zxbg I just found this article about ternary and as you said storage needs to be in ternary too. The main reason seems to be it is close to the value of “e” https://www.techopedia.com/why-not-ternary-computers/2/32427 https://www.techopedia.com/why-not-ternary-computers/2/32427
  - roywiggins 9y ago
    Unless you have access to physical trits, you're going to be emulating them with two bits each...
    - jerf 9y ago
      There's actually this really clever emulation you can do where you don't blindly use two bits per trit, but fractionally pack the trits into bits such that you don't end up with any wasted bits. This has the further advantage that since it produces the same inputs and outputs as binary encoding that you can use the native binary hardware on the CPU to do all your math. Because it is the native binary encoding. One of the reasons I wouldn't trust IOTA any farther than I could personally throw the entire development team is that writing a strong cryptocurrency requires a deep understanding of mathematics, and theirs is the sort of shallow I'd expect from a high schooler. They appear to be operating under the misapprehension that numbers come with their base built into them, as if they are operating in a mathematics regime where "10 != 0xa != 012(octal) != 1010(binary)", and anyone that confused about mathematics is not someone you want designing your CSS stylesheets ("no, that's rgb(255, 255, 255), you really want #fff. Or maybe #ffffff."), let alone your cryptocurrency.
      - roywiggins 9y ago
        I guess I shouldn't be surprised that you can use a clever encoding to work with trits natively but "clever encodings to pack trits" seems like something of an antifeature when we're talking about anything within a mile of cryptography...
- fabian2k 9y ago
  If I understood it correctly, they're claiming that it's more efficient on native ternary hardware, and it has no advantage on binary hardware. The people behind IOTA are also connected in some fashion to the development of such a ternary computer.
  - jebeng 9y ago
    Oh wow. Now that you mention this I totally remember this project from like 2015 when some people, including Come-From-Beyond, were raising money for some sort of ternary computing hardware project. Totally forgot about it, and never put two and two together(three and three? :P) to link it to IOTA. Now that they are quite wealthy, I wonder if it will become a reality. I'm highly skeptical, but I really don't know enough to form a real opinion. Definitely one of those things that make you go 'huh?' though.
dsacco 9y ago
Ternary computation does reduce the number of operations for things like multiplication, that’s correct. But if the rest of the modern world isn’t also ternary, you actually introduce performance overhead that reduces your overall efficiency. A bit-based computation has two states, and a trit-based computation has three. Since 3 isn’t a power of 2 (or even, for that matter), you’ll have a remainder when you convert your ternary multiplication to binary. Considering this, it doesn’t really make sense that the Iota founders made a decision to not only introduce a novel cryptocurrency system (the tangle), but to also try to push ternary adoption. That’s a monolithic amount of work from a research and business perspective, and it strains credulity that they could engage in this with a full understanding of what they’re doing. Their work is completely divorced from the rest of the academic community, which has many well-respected researchers with a vetted track record already. Innovation usually happens in incremental steps forward, building from the (recent) work of others. Much more rarely, it manifests as a large leap forward without significant cumulative work. But it approximately never happens that a new state of the art combines several large and orthogonal leaps forward. It’s not impossible, but my outlook is strongly pesimisstic. When you combine this with the frankly odd (and potentially unethical) behavior the founders are showing - rolling their own cryptography out of ignorance or to attempt to sabotage other open source projects - it becomes extremely hard to take the work seriously.

imglorp 9y ago

> 8 times more efficient than binary They're claiming this, for what operation? And similar question for the 10% usage claim. I suspect storage will be the same when you get done packing into machine words and storing those, yes? Unless they're not using all the bits in a word... Any independent references for ternary efficiencies for any operation?

smrtfckr 9y ago
IOTA is built for the internet of things and more specifically for sensors operating in the field. Perhaps with limited acces, perhaps with long intervals between service periods. Generally, you would't care about an 8 fold efficiency increase when you're hooked up onto a powerline but remote sensing equipment does.
- imglorp 9y ago
  Great, someone else quoting 8x efficiency. Maybe you know about the claim? We understand the motivation, which keeps getting mentioned, but what I think we all want to know is: 8x efficiency _of_what_compared_to_what_?
- PinguTS 9y ago
  Such devices are using ARM Cortex M0, M1, or M3 processors. Maybe even lower power Atmel stuff like in the Arduino. Such devices are connected via either expansive uplinks or slow uplinks like with LPWAN. For all these applications IOTA is not feasible. It is not feasible because the ternary logic needs to be emulated on a processor that itself has not even enough beef to do cryptographic functions by itself. If they have cryptographic support like the e.g. the SE series from STM with a Cortex M0 as used in the Ledger Nano S, then this is optimized for standard cryptographic algorithms and not for some newly invented stuff. Also if we talk about IOT sensor networking with LPWAN (LoRa, SigFox, ...) than bandwith is a real issue. That means collecting transactions from others and doing PoW on that before sending of your own transaction is completely unrealistic. The remote sensor just wants to transmit its data. But that would mean the LPWAN Gateway needs to be a full node doing the PoW. But LPWAN works, that everybody can listen to transaction request and so we are back to the collisions problem.
etunity 9y ago
I am not sure about 8 times efficiency and 10% argument. The founder said that in his interview yesterday. https://m.youtube.com/watch?v=GwhJQ67zxbg https://m.youtube.com/watch?v=GwhJQ67zxbg I just found this article about ternary and as you said storage needs to be in ternary too. The main reason seems to be it is close to the value of “e” https://www.techopedia.com/why-not-ternary-computers/2/32427 https://www.techopedia.com/why-not-ternary-computers/2/32427
- roywiggins 9y ago
  Unless you have access to physical trits, you're going to be emulating them with two bits each...
  - jerf 9y ago
    There's actually this really clever emulation you can do where you don't blindly use two bits per trit, but fractionally pack the trits into bits such that you don't end up with any wasted bits. This has the further advantage that since it produces the same inputs and outputs as binary encoding that you can use the native binary hardware on the CPU to do all your math. Because it is the native binary encoding. One of the reasons I wouldn't trust IOTA any farther than I could personally throw the entire development team is that writing a strong cryptocurrency requires a deep understanding of mathematics, and theirs is the sort of shallow I'd expect from a high schooler. They appear to be operating under the misapprehension that numbers come with their base built into them, as if they are operating in a mathematics regime where "10 != 0xa != 012(octal) != 1010(binary)", and anyone that confused about mathematics is not someone you want designing your CSS stylesheets ("no, that's rgb(255, 255, 255), you really want #fff. Or maybe #ffffff."), let alone your cryptocurrency.
    - roywiggins 9y ago
      I guess I shouldn't be surprised that you can use a clever encoding to work with trits natively but "clever encodings to pack trits" seems like something of an antifeature when we're talking about anything within a mile of cryptography...
fabian2k 9y ago
If I understood it correctly, they're claiming that it's more efficient on native ternary hardware, and it has no advantage on binary hardware. The people behind IOTA are also connected in some fashion to the development of such a ternary computer.
- jebeng 9y ago
  Oh wow. Now that you mention this I totally remember this project from like 2015 when some people, including Come-From-Beyond, were raising money for some sort of ternary computing hardware project. Totally forgot about it, and never put two and two together(three and three? :P) to link it to IOTA. Now that they are quite wealthy, I wonder if it will become a reality. I'm highly skeptical, but I really don't know enough to form a real opinion. Definitely one of those things that make you go 'huh?' though.

dsacco 9y ago

Ternary computation does reduce the number of operations for things like multiplication, that’s correct. But if the rest of the modern world isn’t also ternary, you actually introduce performance overhead that reduces your overall efficiency. A bit-based computation has two states, and a trit-based computation has three. Since 3 isn’t a power of 2 (or even, for that matter), you’ll have a remainder when you convert your ternary multiplication to binary. Considering this, it doesn’t really make sense that the Iota founders made a decision to not only introduce a novel cryptocurrency system (the tangle), but to also try to push ternary adoption. That’s a monolithic amount of work from a research and business perspective, and it strains credulity that they could engage in this with a full understanding of what they’re doing. Their work is completely divorced from the rest of the academic community, which has many well-respected researchers with a vetted track record already. Innovation usually happens in incremental steps forward, building from the (recent) work of others. Much more rarely, it manifests as a large leap forward without significant cumulative work. But it approximately never happens that a new state of the art combines several large and orthogonal leaps forward. It’s not impossible, but my outlook is strongly pesimisstic. When you combine this with the frankly odd (and potentially unethical) behavior the founders are showing - rolling their own cryptography out of ignorance or to attempt to sabotage other open source projects - it becomes extremely hard to take the work seriously.