mably · January 25, 2016 17:44
diff --git a/gistfile1.txt b/gistfile1.txt
 09:52:49 <Tasoshi> I'm trying to understand the weighing of 4:1. If that is increased to 6:1 for example, does that mean an effective size of 6MB can be achieved?
 10:03:23 → proslogion and AaronvanW joined  
 10:07:36 <•CodeShark> Tasoshi: that would be a worse case attack scenario
 10:08:06 <•CodeShark> worst case attack scenario
 10:08:43 <•CodeShark> "Effective size"is typical usage
 10:09:01 <•CodeShark> Different kinds of transactions get different discounts
 10:16:09 <Tasoshi> So if the discount is increased to something crazy, lets say 50:1, does that mean that in a worst case attack scenario a size of 50MB can be reached? What do you mean by worst case attack scenario by the way?
 10:17:41 <•CodeShark> It means it's an absurd spam scenario
 10:18:07 <•CodeShark> someone just filling blocks with useless stuff
 10:19:17 <•CodeShark> An expensive attack...but not out-of-the-question
 10:20:31 ⇐ cbit and ThomasV quit  
 10:25:27 <Tasoshi> Ah, I see. What do you mean by discount?
 10:30:01 <•CodeShark> The cost reduction compared to the current limit as percentage of available space
 10:31:00 <Tasoshi> by cost reduction do you mean the fee that you have to pay?
 10:31:43 <Tasoshi> I'm just reading this and trying to understand it in plain language: "Blocks are currently limited to 1,000,000 bytes (1MB) total size. We change this restriction as follows:
 10:31:44 <Tasoshi> Block cost is defined. The cost of each byte in the existing header and transactions is 4, while the cost of each byte in witness data is 1.
 10:31:44 <Tasoshi> The new rule is total block cost ≤ 4,000,000. "
 10:32:06 <•CodeShark> i.e. a transaction where half the data is witness data would cost 0.5 + 4*0.5 = 2.5 cost units per byte
 10:32:46 <Tasoshi> Is this suggesting that in effect there would no longer be a maxblocksize limit, but this would instead be replaced with a block cost?
 10:32:54 <•CodeShark> yes
 10:35:53 <Tasoshi> I see, so if we increase the cost of header data then we can include more transactions assuming the witness data remains the same?
 10:36:23 → ThomasV joined (~ThomasV@unaffiliated/thomasv)
 10:38:08 <Tasoshi> And going back to your equation, what does 2.5 cost units mean? Is that 2.5 satoshis per byte?
 10:43:25 <•CodeShark> No - it means we can fit 4,000,000 / 2.5 such transactions per block
 10:43:53 <•CodeShark> I mean
 10:44:18 <•CodeShark> 4,000,000 / (2.5*tx_bytes)
 10:45:01 <Tasoshi> Oh, I see, so 2.5 MB?
 10:49:15 <•CodeShark> the formula would be (b + 4w) / (b + w)
 10:49:44 <•CodeShark> Where b is the nonwitness data size and w is the witness data size
 10:49:55 <•CodeShark> For a 4:1 discount
 10:52:03 <•CodeShark> So a transaction where a third of its data is witness data gets throughput increase (1/3 + 4/3) = 5/3
 10:52:25 <•CodeShark> err
 10:52:32 <•CodeShark> Sorry, messed that one up
 10:53:01 <•CodeShark> 2/3 is witness data
 10:53:31 <•CodeShark> 1/3 + 8/3 = 3
 10:53:52 <Tasoshi> I'm not understanding what a cost unit is. Is that a fee, is that the size? So in your example formula earlier where the equation equals 2.5. When we are about to have the transaction, what do we do with this number (2.5)?
 10:54:13 <•CodeShark> trying to type on mobile - sorry :)
 10:54:24 <Tasoshi> no worries
 10:55:44 <•CodeShark> So in the example of 1/2 the tx being witness data, we have 4,000,000 / 2.5 = 1,600,000
 10:55:58 <•CodeShark> Which means an effective block size of 1.6
 10:56:01 <•CodeShark> Mb
 10:56:04 <Tasoshi> Ohhh, 2.5 is the discount?
 10:56:18 <•CodeShark> 2.5 is the cost per byte
 10:56:49 <•CodeShark> Each block has a limited amount of resources
 10:57:12 <•CodeShark> The cost is how much of that finite resource is consumed
 10:57:47 <Tasoshi> How is the cost paid?
 10:57:56 <•CodeShark> free market
 10:58:06 <•CodeShark> Miners can pick which transactions to include
 10:58:15 <Tasoshi> I mean, in the formula.
 10:59:07 <•CodeShark> The cost is not a measure of how much the sender pays - the fee is determined by supply and demand
 10:59:57 <•CodeShark> at least assuming economically rational miners
 11:00:33 <•CodeShark> Supply is limited to 4,000,000 cost units per block
 11:02:13 <•CodeShark> I think the most useful number for comparison is the throughput multiplier
 11:03:38 <Tasoshi> I guess I am confused by the word "cost unit" and I think you mentioned discount earlier
 11:04:31 <•CodeShark> yeah - we're still formalizing the vocabulary ;)
 11:04:55 <•CodeShark> "Discount" is the ratio of nonwitness cost to witness cost
 11:05:25 <•CodeShark> "Cost" is how much of the 4,000,000 per block is consumed
 11:05:53 <Tasoshi> I see, so cost means byte?
 11:06:46 <•CodeShark> it's 4b + w
 11:07:07 <Tasoshi> What's w?
 11:07:51 <•CodeShark> b is size of the nonwitness data (the base data) in bytes, w is the size of the witness data in bytes
 11:08:17 <•CodeShark> I mistyped some of the formulas above - damn touch keyboard
 11:09:06 <•CodeShark> The most interesting formula is probably TM = 4(b + w) / (4b + w)
 11:09:21 <•CodeShark> The throughput multiplier
 11:09:57 <•CodeShark> This is the factor by which # of transactions per block is increased
 11:11:07 <•CodeShark> So a transaction where 2/3 is witness data gets a TM of 4 / (4/3 + 2/3) = 2
 11:11:36 <Tasoshi> 2MB?
 11:12:00 <•CodeShark> 2MB would be the "effective block size"
 11:12:44 <•CodeShark> The effective block size is the size a block would have to be if we have no witness discount
 11:13:38 <•CodeShark> The effective block size is TM megabytes
 11:13:55 <•CodeShark> Make sense?
 11:14:42 <Tasoshi> Sort of, so what is the actual blocksize once the witness discount is included in your recent example?
 11:20:23 <Tasoshi> Or are they both the same at 2MB?
 11:21:13 <•CodeShark> yes
 11:25:06 <Tasoshi> So, the effective blocksize is the same as the actual size of a block?
 11:25:32 → Thireus joined ([email protected])
 11:25:43 <Thireus> hey guys, what's up
 11:26:25 <•CodeShark> Hi Thireus
 11:26:49 <•CodeShark> Tasoshi: the actual size of the block depends on all the transactions in it
 11:27:42 <•CodeShark> and might be below the max
 11:27:58 <•CodeShark> we should actually call it effective max block size
 11:30:26 <Tasoshi> Sorry, I'm just confused because you said: "The effective block size is the size a block would have to be if we have no witness discount". So, presumably, if we include the witness discount the size would be different? I was wondering what size that would be as per your example
 11:32:23 <Tasoshi> not sure if I am understanding correctly, but I used 0.33 + 4*0.33 = 1.65 | 4000000 / 1.65= 2424242
 11:32:38 <•CodeShark> ok, so if 2/3 of transaction data is witness data, it means the effective max block size is 2MB - this means if we had no witness discount, we would get the same throughput as if we had a 2MB max block size
 11:33:34 <•CodeShark> it should be 4*0.33 + 0.66
 11:33:51 <•CodeShark> 4b + w
 11:36:28 <Tasoshi> Oh, I see, you saying it would be same as if maxblocksize=2mb, right
 11:38:11 <Tasoshi> So, if 2/3 of witness data gives an effective size of 2mb, then even if it is 100% witness data we would not reach 4mb?
 11:39:32 <•CodeShark> No, if it's 100% witness data we get 4MB...but that's an attack scenario
 11:39:48 <•CodeShark> More typical is 66% witness data
 11:40:02 <Tasoshi> Oh yeah, because 4x0+1
 11:45:08 → nullpt joined ([email protected])
 11:58:38 <Tasoshi> So let me see if I have understood. Supposing we have a transaction with 500 bytes in non witness data and 700 bytes in witness data. Do we just use the formula of 4x500 +700 = 2700 | 4000000/2700 = 1481 bytes - So that's the maxblocksize when the actual size is 1200 bytes?
 12:04:44 <•CodeShark> 4(500 + 700) / (4*500 + 700) = 4(1200) / (2700) = 4800/2700
 12:05:03 <Tasoshi> I am not sure I have understood however because if we have a transaction or a set of transactions with 2700 bytes and only 100 bytes of witness data we get 4 x 2700 +100 = 10900 | 4m/10900=366
 12:05:32 <•CodeShark> 4800/2700 ~ 1.77
 12:05:51 <•CodeShark> that's thr throughput multiplier
 12:07:54 <Tasoshi> What do we then do with the number you just provided, 1.77?
 12:10:05 → aschildbach joined  ⇐ nullpt__ quit  
 12:26:20 <•CodeShark> It means 1.77x increase in transaction throughput
 12:26:51 <•CodeShark> 1.77MB effective max block size
	09:52:49 <Tasoshi> I'm trying to understand the weighing of 4:1. If that is increased to 6:1 for example, does that mean an effective size of 6MB can be achieved?
	10:03:23 → proslogion and AaronvanW joined
	10:07:36 <•CodeShark> Tasoshi: that would be a worse case attack scenario
	10:08:06 <•CodeShark> worst case attack scenario
	10:08:43 <•CodeShark> "Effective size"is typical usage
	10:09:01 <•CodeShark> Different kinds of transactions get different discounts
	10:16:09 <Tasoshi> So if the discount is increased to something crazy, lets say 50:1, does that mean that in a worst case attack scenario a size of 50MB can be reached? What do you mean by worst case attack scenario by the way?
	10:17:41 <•CodeShark> It means it's an absurd spam scenario
	10:18:07 <•CodeShark> someone just filling blocks with useless stuff
	10:19:17 <•CodeShark> An expensive attack...but not out-of-the-question
	10:20:31 ⇐ cbit and ThomasV quit
	10:25:27 <Tasoshi> Ah, I see. What do you mean by discount?
	10:30:01 <•CodeShark> The cost reduction compared to the current limit as percentage of available space
	10:31:00 <Tasoshi> by cost reduction do you mean the fee that you have to pay?
	10:31:43 <Tasoshi> I'm just reading this and trying to understand it in plain language: "Blocks are currently limited to 1,000,000 bytes (1MB) total size. We change this restriction as follows:
	10:31:44 <Tasoshi> Block cost is defined. The cost of each byte in the existing header and transactions is 4, while the cost of each byte in witness data is 1.
	10:31:44 <Tasoshi> The new rule is total block cost ≤ 4,000,000. "
	10:32:06 <•CodeShark> i.e. a transaction where half the data is witness data would cost 0.5 + 4*0.5 = 2.5 cost units per byte
	10:32:46 <Tasoshi> Is this suggesting that in effect there would no longer be a maxblocksize limit, but this would instead be replaced with a block cost?
	10:32:54 <•CodeShark> yes
	10:35:53 <Tasoshi> I see, so if we increase the cost of header data then we can include more transactions assuming the witness data remains the same?
	10:36:23 → ThomasV joined (~ThomasV@unaffiliated/thomasv)
	10:38:08 <Tasoshi> And going back to your equation, what does 2.5 cost units mean? Is that 2.5 satoshis per byte?
	10:43:25 <•CodeShark> No - it means we can fit 4,000,000 / 2.5 such transactions per block
	10:43:53 <•CodeShark> I mean
	10:44:18 <•CodeShark> 4,000,000 / (2.5*tx_bytes)
	10:45:01 <Tasoshi> Oh, I see, so 2.5 MB?
	10:49:15 <•CodeShark> the formula would be (b + 4w) / (b + w)
	10:49:44 <•CodeShark> Where b is the nonwitness data size and w is the witness data size
	10:49:55 <•CodeShark> For a 4:1 discount
	10:52:03 <•CodeShark> So a transaction where a third of its data is witness data gets throughput increase (1/3 + 4/3) = 5/3
	10:52:25 <•CodeShark> err
	10:52:32 <•CodeShark> Sorry, messed that one up
	10:53:01 <•CodeShark> 2/3 is witness data
	10:53:31 <•CodeShark> 1/3 + 8/3 = 3
	10:53:52 <Tasoshi> I'm not understanding what a cost unit is. Is that a fee, is that the size? So in your example formula earlier where the equation equals 2.5. When we are about to have the transaction, what do we do with this number (2.5)?
	10:54:13 <•CodeShark> trying to type on mobile - sorry :)
	10:54:24 <Tasoshi> no worries
	10:55:44 <•CodeShark> So in the example of 1/2 the tx being witness data, we have 4,000,000 / 2.5 = 1,600,000
	10:55:58 <•CodeShark> Which means an effective block size of 1.6
	10:56:01 <•CodeShark> Mb
	10:56:04 <Tasoshi> Ohhh, 2.5 is the discount?
	10:56:18 <•CodeShark> 2.5 is the cost per byte
	10:56:49 <•CodeShark> Each block has a limited amount of resources
	10:57:12 <•CodeShark> The cost is how much of that finite resource is consumed
	10:57:47 <Tasoshi> How is the cost paid?
	10:57:56 <•CodeShark> free market
	10:58:06 <•CodeShark> Miners can pick which transactions to include
	10:58:15 <Tasoshi> I mean, in the formula.
	10:59:07 <•CodeShark> The cost is not a measure of how much the sender pays - the fee is determined by supply and demand
	10:59:57 <•CodeShark> at least assuming economically rational miners
	11:00:33 <•CodeShark> Supply is limited to 4,000,000 cost units per block
	11:02:13 <•CodeShark> I think the most useful number for comparison is the throughput multiplier
	11:03:38 <Tasoshi> I guess I am confused by the word "cost unit" and I think you mentioned discount earlier
	11:04:31 <•CodeShark> yeah - we're still formalizing the vocabulary ;)
	11:04:55 <•CodeShark> "Discount" is the ratio of nonwitness cost to witness cost
	11:05:25 <•CodeShark> "Cost" is how much of the 4,000,000 per block is consumed
	11:05:53 <Tasoshi> I see, so cost means byte?
	11:06:46 <•CodeShark> it's 4b + w
	11:07:07 <Tasoshi> What's w?
	11:07:51 <•CodeShark> b is size of the nonwitness data (the base data) in bytes, w is the size of the witness data in bytes
	11:08:17 <•CodeShark> I mistyped some of the formulas above - damn touch keyboard
	11:09:06 <•CodeShark> The most interesting formula is probably TM = 4(b + w) / (4b + w)
	11:09:21 <•CodeShark> The throughput multiplier
	11:09:57 <•CodeShark> This is the factor by which # of transactions per block is increased
	11:11:07 <•CodeShark> So a transaction where 2/3 is witness data gets a TM of 4 / (4/3 + 2/3) = 2
	11:11:36 <Tasoshi> 2MB?
	11:12:00 <•CodeShark> 2MB would be the "effective block size"
	11:12:44 <•CodeShark> The effective block size is the size a block would have to be if we have no witness discount
	11:13:38 <•CodeShark> The effective block size is TM megabytes
	11:13:55 <•CodeShark> Make sense?
	11:14:42 <Tasoshi> Sort of, so what is the actual blocksize once the witness discount is included in your recent example?
	11:20:23 <Tasoshi> Or are they both the same at 2MB?
	11:21:13 <•CodeShark> yes
	11:25:06 <Tasoshi> So, the effective blocksize is the same as the actual size of a block?
	11:25:32 → Thireus joined ([email protected])
	11:25:43 <Thireus> hey guys, what's up
	11:26:25 <•CodeShark> Hi Thireus
	11:26:49 <•CodeShark> Tasoshi: the actual size of the block depends on all the transactions in it
	11:27:42 <•CodeShark> and might be below the max
	11:27:58 <•CodeShark> we should actually call it effective max block size
	11:30:26 <Tasoshi> Sorry, I'm just confused because you said: "The effective block size is the size a block would have to be if we have no witness discount". So, presumably, if we include the witness discount the size would be different? I was wondering what size that would be as per your example
	11:32:23 <Tasoshi> not sure if I am understanding correctly, but I used 0.33 + 4*0.33 = 1.65 \| 4000000 / 1.65= 2424242
	11:32:38 <•CodeShark> ok, so if 2/3 of transaction data is witness data, it means the effective max block size is 2MB - this means if we had no witness discount, we would get the same throughput as if we had a 2MB max block size
	11:33:34 <•CodeShark> it should be 4*0.33 + 0.66
	11:33:51 <•CodeShark> 4b + w
	11:36:28 <Tasoshi> Oh, I see, you saying it would be same as if maxblocksize=2mb, right
	11:38:11 <Tasoshi> So, if 2/3 of witness data gives an effective size of 2mb, then even if it is 100% witness data we would not reach 4mb?
	11:39:32 <•CodeShark> No, if it's 100% witness data we get 4MB...but that's an attack scenario
	11:39:48 <•CodeShark> More typical is 66% witness data
	11:40:02 <Tasoshi> Oh yeah, because 4x0+1
	11:45:08 → nullpt joined ([email protected])
	11:58:38 <Tasoshi> So let me see if I have understood. Supposing we have a transaction with 500 bytes in non witness data and 700 bytes in witness data. Do we just use the formula of 4x500 +700 = 2700 \| 4000000/2700 = 1481 bytes - So that's the maxblocksize when the actual size is 1200 bytes?
	12:04:44 <•CodeShark> 4(500 + 700) / (4*500 + 700) = 4(1200) / (2700) = 4800/2700
	12:05:03 <Tasoshi> I am not sure I have understood however because if we have a transaction or a set of transactions with 2700 bytes and only 100 bytes of witness data we get 4 x 2700 +100 = 10900 \| 4m/10900=366
	12:05:32 <•CodeShark> 4800/2700 ~ 1.77
	12:05:51 <•CodeShark> that's thr throughput multiplier
	12:07:54 <Tasoshi> What do we then do with the number you just provided, 1.77?
	12:10:05 → aschildbach joined ⇐ nullpt__ quit
	12:26:20 <•CodeShark> It means 1.77x increase in transaction throughput
	12:26:51 <•CodeShark> 1.77MB effective max block size