jl777
Legendary
Offline
Activity: 1176
Merit: 1134
|
|
February 27, 2014, 03:51:34 PM |
|
I proposed a simpler method, but CIYAM said it was impossible. Do you think using current method (without the random factor) we can simply reduce the time between blocks to 50 seconds? 30 seconds? 10 seconds?
James
We can reduce. But if u set gap between blocks to 10 sec then u'll need 6 times more confirmations to get the same reliability. So with current internet latency the 60 seconds confirmation time is the fastest practical time? What if we reduce the network to just hallmarked nodes all with high speed internet? What is the fastest practical time to get current reliability (just among the hallmarked nodes)? James
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 03:51:56 PM |
|
Would Nxt be slower or faster?
Put it this way: 1 x 1 minute confirmation takes 1 minute and 6 x 10 second confirmations takes 1 minute. Both are most likely about the same level of security.
|
|
|
|
redsn0w
Legendary
Offline
Activity: 1778
Merit: 1043
#Free market
|
|
February 27, 2014, 03:53:18 PM |
|
So... more nodes ?
It doesn't *change* the *latency* to have more nodes. The latency is due to the physical hardware of the internet itself and some of the software (in particular things like the GCF) that sit at the fairly low levels above that. Adding "more hops" can actually only make things *slower*. then What's the solution ?
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 03:53:55 PM |
|
What if we reduce the network to just hallmarked nodes all with high speed internet? What is the fastest practical time to get current reliability (just among the hallmarked nodes)?
It only helps the "hallmarked nodes" in "talking to each other" - it doesn't change the time it takes your own node (if you are a retailer for example) to talk to them. So sure you could have those nodes probably do things much faster "between themselves" but still the "outsider" can't catch up.
|
|
|
|
Come-from-Beyond
Legendary
Offline
Activity: 2142
Merit: 1010
Newbie
|
|
February 27, 2014, 03:53:59 PM |
|
Would Nxt be slower or faster?
Depends on topology.
|
|
|
|
ChuckOne
Sr. Member
Offline
Activity: 364
Merit: 250
☕ NXT-4BTE-8Y4K-CDS2-6TB82
|
|
February 27, 2014, 03:55:04 PM |
|
But for simple transactions?
That's where we have to work out "trade offs". IMO having lots of forks is not a great idea (a fork of > 2 rarely happens in Bitcoin). Remember the "average Joe" going into a 7-11 to purchase a "can of soda" probably isn't even "capable" of trying to do a "double spend" so is it any more likely to occur than if they "slipped the soda can into their pocket"? Not sure why what trade-offs exactly you are talking about. -------------------------- Btw: For the can of soda aka shopping, we would need a protocol like this: 1) Joe creates a transaction with the parameters specified by the supermarket. (in fact some nice credit card looking object will do) 2) Joe sends the transaction data to the supermarket and get the can in exchange 3) The supermarket would be eager to send that transaction to the network ASAP 4) => supermarket will even run its own node to secure their transactions. If the transaction was send to a fork node, it can be resend over and over again until the supermarket get its money. Something wrong with that?
|
|
|
|
bitcoinpaul
|
|
February 27, 2014, 03:56:15 PM |
|
image
I lold very hard I'm still lolling. My dog thinks I'm stupid. And I'm hodling.
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 03:56:30 PM |
|
Would Nxt be slower or faster?
Depends on topology. Again - I think CfB and I are "in tune" with this. You can't improve your own latency issues by having "some other servers" talk faster to each other. At the end of the day your "slow connection" simply "can't keep up" so it won't be able to see those fast flying confirmations and more than likely will end up on a fork.
|
|
|
|
ChuckOne
Sr. Member
Offline
Activity: 364
Merit: 250
☕ NXT-4BTE-8Y4K-CDS2-6TB82
|
|
February 27, 2014, 03:58:28 PM |
|
I proposed a simpler method, but CIYAM said it was impossible. Do you think using current method (without the random factor) we can simply reduce the time between blocks to 50 seconds? 30 seconds? 10 seconds?
James
We can reduce. But if u set gap between blocks to 10 sec then u'll need 6 times more confirmations to get the same reliability. So with current internet latency the 60 seconds confirmation time is the fastest practical time? What if we reduce the network to just hallmarked nodes all with high speed internet? What is the fastest practical time to get current reliability (just among the hallmarked nodes)? James Not sure why 60 seconds should be the fastest practical time. Simple transactions as buying a can of soda do not require 1440 blocks. As long as the "supermarket" caches its precious not-confirmed transactions until they are, nobody's in trouble. @CfB Could you elaborate?
|
|
|
|
bitcoinpaul
|
|
February 27, 2014, 04:00:09 PM |
|
Quick question: Does local signing change anything of the soda-can situation?
edit: Ok. Ignore.
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 04:00:22 PM |
|
If the transaction was send to a fork node, it can be resend over and over again until the supermarket get its money.
Not if the balance of said account has become *zero* in the meantime due to the supermarket not having fast enough internet to see the other tx that was sent by "tricky Joe" at the same time he bought the soda from another location (which emptied his balance via a higher speed route) as he had set that up perhaps via a VPS or some other way (as I said "average Joe" would be very unlikely to be able to get away with this).
|
|
|
|
jl777
Legendary
Offline
Activity: 1176
Merit: 1134
|
|
February 27, 2014, 04:01:26 PM |
|
What if we reduce the network to just hallmarked nodes all with high speed internet? What is the fastest practical time to get current reliability (just among the hallmarked nodes)?
It only helps the "hallmarked nodes" in "talking to each other" - it doesn't change the time it takes your own node (if you are a retailer for example) to talk to them. So sure you could have those nodes probably do things much faster "between themselves" but still the "outsider" can't catch up. Please try to understand what I am saying. You agree that we can have hallmarked nodes going at a much faster speed. I understand that it will only be between themselves. Now, lets assume that we place the hallmarked nodes in a topology such that with the actual Internet of today, all nodes are a short ping distance away from at least one hallmarked node. And we restrict non-hallmarked nodes to only use peers with a fast enough ping time. This two tiered approach is kind of what is going on now, but not strictly. I cant understand why the overall processing wont go any faster if all nodes are fast to a hallmark and all hallmarks are fast to each other. So, in China there would be hallmarked servers that you can ping in 50 milliseconds. We make sure the hallmarks in China have a fast path outside. Are you saying that with this structure we are still at 60 seconds? If so, math please James
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 04:01:57 PM |
|
Quick question: Does local signing change anything of the soda-can situation?
No - you can have your balance emptied by a *faster* tx (if you know how to achieve that) so that your tx at the supermarket will simply fail due to zero balance.
|
|
|
|
landomata
Legendary
Offline
Activity: 2184
Merit: 1000
|
|
February 27, 2014, 04:02:20 PM |
|
If the transaction was send to a fork node, it can be resend over and over again until the supermarket get its money.
Not if the balance of said account has become *zero* in the meantime due to the supermarket not having fast enough internet to see the other tx that was sent by "tricky Joe" at the same time he bought the soda from another location (which emptied his balance via a higher speed route) as he had set that up perhaps via a VPS or some other way (as I said "average Joe" would be very unlikely to be able to get away with this). Joe is buying a coke he can;t be in two places at once.
|
|
|
|
bitcoinpaul
|
|
February 27, 2014, 04:03:32 PM |
|
If the transaction was send to a fork node, it can be resend over and over again until the supermarket get its money.
Not if the balance of said account has become *zero* in the meantime due to the supermarket not having fast enough internet to see the other tx that was sent by "tricky Joe" at the same time he bought the soda from another location (which emptied his balance via a higher speed route) as he had set that up perhaps via a VPS or some other way (as I said "average Joe" would be very unlikely to be able to get away with this). Joe is buying a coke he can;t be in two places at once. If Joe has no friends, it is not possible
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 04:03:46 PM |
|
So, in China there would be hallmarked servers that you can ping in 50 milliseconds. We make sure the hallmarks in China have a fast path outside.
How are you going to "make sure" of this? Are you going to "beat" the GCF? If so please let me know the "secret sauce".
|
|
|
|
CIYAM
Legendary
Offline
Activity: 1890
Merit: 1086
Ian Knowles - CIYAM Lead Developer
|
|
February 27, 2014, 04:05:56 PM |
|
Joe is buying a coke he can;t be in two places at once.
If Joe has no friends, it is not possible If he has a suitable AT then he might not need a friend.
|
|
|
|
landomata
Legendary
Offline
Activity: 2184
Merit: 1000
|
|
February 27, 2014, 04:06:28 PM |
|
If Joe has no friends, it is not possible 1) If Joe wanted to steal a coke he could simple walk out of the store without paying. 2) I understand we must act on reality not our limited assumptions....but I'm sure some very smart people said humans would never fly. Seeing the impossible is also genius. EDIT: seeing pass the impossible is genius
|
|
|
|
opticalcarrier
|
|
February 27, 2014, 04:06:39 PM |
|
What do you think about limiting max forging power to anywhere between 1% to 5%
I didn't come to a conclusion yet. i still say this is one of the things we all need to talk through and address first. how do we prevent centralization of forging in a non-trustless manner? and can someone answer a question on the following 2 scenarios: 1) a single account has 91% of all NXT in its balance 2) a single account balance + all the effectiveBalances leased to that single account where the sum is 91% of all NXT Do both these scenarios present the same risk of a 91% attack? It seems to me that they do, but some people seem to have been suggesting otherwise, or at least thats how I understand some conversations Also, can anyone answer this: in what cases would you use broadcastTransaction API? Is it only for light-clients to use to send to a full node? I thought we were saying before that if some forger never picked up your transaction you could use broadcastTransaction to resend it, but that requires full bytes of the transaction, and you cannot obtain that unless you pull it from the blockchain, which means it is already *in* the blockchain.
I dont understand?
|
|
|
|
jl777
Legendary
Offline
Activity: 1176
Merit: 1134
|
|
February 27, 2014, 04:10:08 PM |
|
So, in China there would be hallmarked servers that you can ping in 50 milliseconds. We make sure the hallmarks in China have a fast path outside.
How are you going to "make sure" of this? Are you going to "beat" the GCF? If so please let me know the "secret sauce". So the "only" problem is that we cant ensure a fast path out of China? We can get the 50 milliseconds ping inside China? If so, I might have a solution, just a matter of cost. Is is fair to assume that we can estimate a block generation time being limited by 200 times the worst ping times of nodes? James
|
|
|
|
|