Variable Block Size Proposal
4 messages, 29 Aug 2015. Started by Justin M. Wray.
Justin M. Wray (2), Jameson Lopp, Jorge Timón
Hey Bitcoiners!
While I am an avid Bitcoin supporter, long-term user, and have done development work on tools and platforms surrounding Bitcoin, I have been very busy these past few weeks and haven't had a chance to fully (or closely) monitor the Block Size debate.
I'm familiar with the basics, and have read abstracts about the front-running proposals (BIP 100, 101, and 102). Though I've honestly not read those in depth either. With that said, I was driving the other day and thought of a potential idea. I'll be clear, this is just an idea, and I haven't fully fleshed it out. But I thought I'd throw it out there and see what people thought.
My Goal:
Provide a variable block size that provides for sustainable, long-term growth, and balances the block propagation, while also being mindful of potential spam attacks.
The Proposal:
Every 2016 blocks (approximately every two weeks, at the same time the difficulty is adjusted), the new block size parameters are calculated.
The calculation determines the average (mean) size of the past 2016 blocks. This "average" size is then doubled (200%) and used as the maximum block size for the subsequent 2016 blocks. At any point, if the new maximum size is calculated to be below 1MB, 1MB is used instead (which prevents regression from our current state).
Introduce a block minimum, the minimum will be 25% of the current maximum, calculated at the same time (that is, every 2016 blocks, at the same time the maximum is calculated). All blocks must be at least this size in order to be valid, for blocks that do not have enough transactions to meet the 25%, padding will be used. This devalues the incentive to mine empty blocks in either an attempt to deflate the block size, or to obtain a propagation advantage. Miners will be incentivized to include transactions, as the block must meet the minimum. This should ensure that even miners wishing to always mine the minimum are still confirming Bitcoin transactions.
At the block in which this is introduced the maximum would stay at 1MB for the subsequent 2016 blocks. With the minimum being enforced of 256KB .
Example:
* Average Block Size for the last 2016 blocks: 724KB
* New Maximum: 1448KB
* New Minimum: 362KB
Example: (Regression Prevention)
* Average Block Size for the last 2016 blocks: 250KB
* New Maximum: 1MB
* New Minimum: 256KB
The Future:
I believe that the 1MB regression prevention might need to be changed in the future, to prevent a large mining population from continually deflating the block size (and keeping us at the 1MB limit).
For this, the hard limit could be changed in the future manually, through a process similar to the current one, though hopefully with far less urgency and hysteria.
Another option is to add an additional calculation, preventing the new maximum from being lower than 75% of the current maximum. This would substantially slow down a block-size deflation attack.
Example of Block-Size Deflation Attack Prevention:
* Average Block Size for the last 2016 blocks: 4MB * New Maximum: 8MB * New Minimum: 2MB
* Average Block Size for the last 2016 blocks: 2MB * New Maximum: 6MB (2 * 200% = 4, 4< 75% of 8, So use 8 * .75 = 6) * New Minimum: 1.5MB
This would provide a maximum growth of 200% per recalculation, but a maximum shrinkage of 75%.
Request For Comments:
I'd love to hear your thoughts. Why wouldn't this work? What portion is flawed? Will the miners support such a proposal? Would this even solve the block size issue?
I will note that I don't find the 100% and 25% to be hard and fast in my idea. Those we're just the values that initially jumped out at me. I could easily see the minimum being anything below 50% (above 50% and the network can never adjust to smaller block sizes). I could also see the maximum being anything over 100%. Lastly, if a inflation attack is a valid concern, a hard upper limit could be set (or the historical 32MB limit could remain).
I think the great part about this variable approach is that the network can adjust to address spikes in volume and readjust once those spikes dissipate.
- -- Thanks!
- ----- Justin M. Wray
I don't think you'll find much support for introducing a mandatory minimum block size. It's quite wasteful to "pad" blocks with transactions that the miner is just sending back to themself. If you want to solve the block propagation issue, I'd recommend instead working on O(1) block propagation.
The Bitcoin Relay Network already allows miners to relay blocks much faster: http://bitcoinrelaynetwork.org/
The next step would be getting O(1) block propagation into the Bitcoin protocol. Check out Gavin's proposal: https://gist.github.com/gavinandresen/e20c3b5a1d4b97f79ac2
- Jameson
On Sat, Aug 29, 2015 at 1:41 AM, Justin M. Wray via bitcoin-dev < bitcoin-dev@lists•linuxfoundation•org> wrote:
Quoted text (133 lines)
> -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA512 > > Hey Bitcoiners! > > While I am an avid Bitcoin supporter, long-term user, and have done > development work on tools and platforms surrounding Bitcoin, I have > been very busy these past few weeks and haven't had a chance to fully > (or closely) monitor the Block Size debate. > > I'm familiar with the basics, and have read abstracts about the > front-running proposals (BIP 100, 101, and 102). Though I've honestly > not read those in depth either. With that said, I was driving > the other day and thought of a potential idea. I'll be clear, this is > just an idea, and I haven't fully fleshed it out. But I thought I'd > throw it out there and see what people thought. > > My Goal: > > Provide a variable block size that provides for sustainable, long-term > growth, and balances the block propagation, while also being mindful > of potential spam attacks. > > The Proposal: > > Every 2016 blocks (approximately every two weeks, at the same time the > difficulty is adjusted), the new block size parameters are calculated. > > The calculation determines the average (mean) size of the past 2016 > blocks. This "average" size is then doubled (200%) and used as the > maximum block size for the subsequent 2016 blocks. At any point, if > the new maximum size is calculated to be below 1MB, 1MB is used > instead (which prevents regression from our current state). > > Introduce a block minimum, the minimum will be 25% of the current > maximum, calculated at the same time (that is, every 2016 blocks, at > the same time the maximum is calculated). All blocks must be at least > this size in order to be valid, for blocks that do not have enough > transactions to meet the 25%, padding will be used. This devalues the > incentive to mine empty blocks in either an attempt to deflate the > block size, or to obtain a propagation advantage. Miners will be > incentivized to include transactions, as the block must meet the > minimum. This should ensure that even miners wishing to always mine > the minimum are still confirming Bitcoin transactions. > > At the block in which this is introduced the maximum would stay at 1MB > for the subsequent 2016 blocks. With the minimum being enforced of 256KB > . > > Example: > > * Average Block Size for the last 2016 blocks: 724KB > * New Maximum: 1448KB > * New Minimum: 362KB > > Example: (Regression Prevention) > > * Average Block Size for the last 2016 blocks: 250KB > * New Maximum: 1MB > * New Minimum: 256KB > > The Future: > > I believe that the 1MB regression prevention might need to be changed > in the future, to prevent a large mining population from continually > deflating the block size (and keeping us at the 1MB limit). > > For this, the hard limit could be changed in the future manually, > through a process similar to the current one, though hopefully with > far less urgency and hysteria. > > Another option is to add an additional calculation, preventing the new > maximum from being lower than 75% of the current maximum. This would > substantially slow down a block-size deflation attack. > > Example of Block-Size Deflation Attack Prevention: > > * Average Block Size for the last 2016 blocks: 4MB > * New Maximum: 8MB > * New Minimum: 2MB > > * Average Block Size for the last 2016 blocks: 2MB > * New Maximum: 6MB (2 * 200% = 4, 4< 75% of 8, So use 8 * .75 = 6) > * New Minimum: 1.5MB > > This would provide a maximum growth of 200% per recalculation, but a > maximum shrinkage of 75%. > > Request For Comments: > > I'd love to hear your thoughts. Why wouldn't this work? What portion > is flawed? Will the miners support such a proposal? Would this even > solve the block size issue? > > I will note that I don't find the 100% and 25% to be hard and fast in > my idea. Those we're just the values that initially jumped out at me. > I could easily see the minimum being anything below 50% (above 50% and > the network can never adjust to smaller block sizes). I could also see > the maximum being anything over 100%. Lastly, if a inflation attack > is a valid concern, a hard upper limit could be set (or the historical > 32MB limit could remain). > > I think the great part about this variable approach is that the > network can adjust to address spikes in volume and readjust once those > spikes dissipate. > > - -- > Thanks! > > - ----- > Justin M. Wray > -----BEGIN PGP SIGNATURE----- > Comment: GPGTools - https://gpgtools.org > > iQIcBAEBCgAGBQJV4UXvAAoJENo/Q5Xwcn83ZWEP/iXAlNk5p9OlOPNSoHkECcxe > AcartxMLrmOvAZVudU4+239TEvwPydmYX/ptmBYgrvRJfm/TWmi0ZbTioxbxTIWM > IlNta1Y8IOHOEgBCtSW01j1PFHIzkBHQGIuqrKHhjcNVGbegXlPm3Da0gjNuTBIe > IV58gf1OfYK2XjuCMQMvo3VyXUKhqbOvBNnZXr+Qo2sAtanmxHQ+TU/gjA02L9LO > bb8WqQDj/veGnMexGh/X58tfQ5KCfLO401F7KnConDaFdKVDikp32zaSXZ7JWf/K > OeseHW1OHHVdYpHvh5VG5GLtYYB5rnq8g7B0/kyx5n4ldB6GkLxzH9CPB0vxpMnZ > dVCS/+EUe/wkHrpRVNhMwP8XfG+8gv9upKg6H/u39XmpL2H2G4cKeot5xRiWRNqY > oJclAeIhDTL1bx/9e/VqvM91ESWpBLs+O8Mh9OzgfbN3gKR6BuoWHNwM9jSMDAT1 > YzwdneSvAEFzgELMlae2QIzAUHno9qkHMkDVbdY3bBtSM9Xz4ditGgnq1D40ZZ+J > zx5WVY7HCebgbk7T35xgKzSKQSEG9zFNW5Dvq66Se3Zpc5vCPw7Q2xwjjPz3zdXQ > Lub0ohVWTzKr05tN1e/nu6keiY5cXRZ0w2MtHb19jtdWyoHEWWHanfOZjgbVSsuA > saFCydA7O4E4BFxgtNze > =JthX > -----END PGP SIGNATURE----- > _______________________________________________ > bitcoin-dev mailing list > bitcoin-dev@lists•linuxfoundation•org > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev >
That's fine too. Obviously the variable maximum would work just fine without a minimum. In fact, with the O(1) propagation proposal, a minimum number of transactions could be enforced, think - a percentage of the current mempool. That's actually far more meaningful to both miners and consumers.
On 8/29/15 10:22 AM, Jameson Lopp wrote:
Quoted text (1 line)
> I don't think you'll find much support for introducing a mandatory minimum block size. It's quite
wasteful to "pad" blocks with transactions that the miner is just sending back to themself. If you want to solve the block propagation issue, I'd recommend instead working on O(1) block propagation.
Quoted text (2 lines)
> > The Bitcoin Relay Network already allows miners to relay blocks much
faster: http://bitcoinrelaynetwork.org/
Quoted text (2 lines)
> > The next step would be getting O(1) block propagation into the Bitcoin
protocol. Check out Gavin's proposal: https://gist.github.com/gavinandresen/e20c3b5a1d4b97f79ac2
Quoted text (4 lines)
> > - Jameson > > On Sat, Aug 29, 2015 at 1:41 AM, Justin M. Wray via bitcoin-dev
<bitcoin-dev@lists•linuxfoundation•org <mailto:bitcoin-dev@lists.linuxfoundation.org>> wrote:
Quoted text (48 lines)
> > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA512 > > Hey Bitcoiners! > > While I am an avid Bitcoin supporter, long-term user, and have done > development work on tools and platforms surrounding Bitcoin, I have > been very busy these past few weeks and haven't had a chance to fully > (or closely) monitor the Block Size debate. > > I'm familiar with the basics, and have read abstracts about the > front-running proposals (BIP 100, 101, and 102). Though I've honestly > not read those in depth either. With that said, I was driving > the other day and thought of a potential idea. I'll be clear, this is > just an idea, and I haven't fully fleshed it out. But I thought I'd > throw it out there and see what people thought. > > My Goal: > > Provide a variable block size that provides for sustainable, long-term > growth, and balances the block propagation, while also being mindful > of potential spam attacks. > > The Proposal: > > Every 2016 blocks (approximately every two weeks, at the same time the > difficulty is adjusted), the new block size parameters are calculated. > > The calculation determines the average (mean) size of the past 2016 > blocks. This "average" size is then doubled (200%) and used as the > maximum block size for the subsequent 2016 blocks. At any point, if > the new maximum size is calculated to be below 1MB, 1MB is used > instead (which prevents regression from our current state). > > Introduce a block minimum, the minimum will be 25% of the current > maximum, calculated at the same time (that is, every 2016 blocks, at > the same time the maximum is calculated). All blocks must be at least > this size in order to be valid, for blocks that do not have enough > transactions to meet the 25%, padding will be used. This devalues the > incentive to mine empty blocks in either an attempt to deflate the > block size, or to obtain a propagation advantage. Miners will be > incentivized to include transactions, as the block must meet the > minimum. This should ensure that even miners wishing to always mine > the minimum are still confirming Bitcoin transactions. > > At the block in which this is introduced the maximum would stay at 1MB > for the subsequent 2016 blocks. With the minimum being enforced of
256KB
Quoted text (84 lines)
> . > > Example: > > * Average Block Size for the last 2016 blocks: 724KB > * New Maximum: 1448KB > * New Minimum: 362KB > > Example: (Regression Prevention) > > * Average Block Size for the last 2016 blocks: 250KB > * New Maximum: 1MB > * New Minimum: 256KB > > The Future: > > I believe that the 1MB regression prevention might need to be changed > in the future, to prevent a large mining population from continually > deflating the block size (and keeping us at the 1MB limit). > > For this, the hard limit could be changed in the future manually, > through a process similar to the current one, though hopefully with > far less urgency and hysteria. > > Another option is to add an additional calculation, preventing the new > maximum from being lower than 75% of the current maximum. This would > substantially slow down a block-size deflation attack. > > Example of Block-Size Deflation Attack Prevention: > > * Average Block Size for the last 2016 blocks: 4MB > * New Maximum: 8MB > * New Minimum: 2MB > > * Average Block Size for the last 2016 blocks: 2MB > * New Maximum: 6MB (2 * 200% = 4, 4< 75% of 8, So use 8 * .75 = 6) > * New Minimum: 1.5MB > > This would provide a maximum growth of 200% per recalculation, but a > maximum shrinkage of 75%. > > Request For Comments: > > I'd love to hear your thoughts. Why wouldn't this work? What portion > is flawed? Will the miners support such a proposal? Would this even > solve the block size issue? > > I will note that I don't find the 100% and 25% to be hard and fast in > my idea. Those we're just the values that initially jumped out at me. > I could easily see the minimum being anything below 50% (above 50% and > the network can never adjust to smaller block sizes). I could also see > the maximum being anything over 100%. Lastly, if a inflation attack > is a valid concern, a hard upper limit could be set (or the historical > 32MB limit could remain). > > I think the great part about this variable approach is that the > network can adjust to address spikes in volume and readjust once those > spikes dissipate. > > - -- > Thanks! > > - ----- > Justin M. Wray > -----BEGIN PGP SIGNATURE----- > Comment: GPGTools - https://gpgtools.org > > iQIcBAEBCgAGBQJV4UXvAAoJENo/Q5Xwcn83ZWEP/iXAlNk5p9OlOPNSoHkECcxe > AcartxMLrmOvAZVudU4+239TEvwPydmYX/ptmBYgrvRJfm/TWmi0ZbTioxbxTIWM > IlNta1Y8IOHOEgBCtSW01j1PFHIzkBHQGIuqrKHhjcNVGbegXlPm3Da0gjNuTBIe > IV58gf1OfYK2XjuCMQMvo3VyXUKhqbOvBNnZXr+Qo2sAtanmxHQ+TU/gjA02L9LO > bb8WqQDj/veGnMexGh/X58tfQ5KCfLO401F7KnConDaFdKVDikp32zaSXZ7JWf/K > OeseHW1OHHVdYpHvh5VG5GLtYYB5rnq8g7B0/kyx5n4ldB6GkLxzH9CPB0vxpMnZ > dVCS/+EUe/wkHrpRVNhMwP8XfG+8gv9upKg6H/u39XmpL2H2G4cKeot5xRiWRNqY > oJclAeIhDTL1bx/9e/VqvM91ESWpBLs+O8Mh9OzgfbN3gKR6BuoWHNwM9jSMDAT1 > YzwdneSvAEFzgELMlae2QIzAUHno9qkHMkDVbdY3bBtSM9Xz4ditGgnq1D40ZZ+J > zx5WVY7HCebgbk7T35xgKzSKQSEG9zFNW5Dvq66Se3Zpc5vCPw7Q2xwjjPz3zdXQ > Lub0ohVWTzKr05tN1e/nu6keiY5cXRZ0w2MtHb19jtdWyoHEWWHanfOZjgbVSsuA > saFCydA7O4E4BFxgtNze > =JthX > -----END PGP SIGNATURE----- > _______________________________________________ > bitcoin-dev mailing list > bitcoin-dev@lists•linuxfoundation•org
<mailto:bitcoin-dev@lists.linuxfoundation.org>
Quoted text (3 lines)
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev > >
- -- Thanks!
- ----- Justin M. Wray
On Sat, Aug 29, 2015 at 4:22 PM, Jameson Lopp via bitcoin-dev <bitcoin-dev@lists•linuxfoundation•org> wrote:
Quoted text (4 lines)
> I don't think you'll find much support for introducing a mandatory minimum > block size. It's quite wasteful to "pad" blocks with transactions that the > miner is just sending back to themself. If you want to solve the block > propagation issue, I'd recommend instead working on O(1) block propagation.
O(1) propagation can never be achieved without a centralized topology. No matter how efficient of an IBLT (or similar) we implement, O(1) can only be achieved for transmitting a block between 2 nodes. The propagation time of a block across the whole network (or even to the miner's sub-network) will always depend on the concrete network topology.
That's not to say that transmitting a block between to peers in constant time wouldn't greatly help with mining centralization concerns related the maximum block size, but I'm concerned about this incorrect "O(1) block propagation" meme keeps spreading. Even with ansibles [1] and safe zk-SNARKS [2] for constant time block validation (somehow removing the trusted setup), both of which are science fiction right now you need to verify the snark proof for every node receiving and relaying the block. At that point block propagation would be meaningless as a miner-centralization concern for not raising the maximum block size though: the minimum CPU costs for being able to mine profitably would be the next concern or "bottleneck".
I completely agree with the minimum block size being inappropriate though. I don't even believe that the stated goals of the size minimum can be accomplished with it.
[1] https://en.wikipedia.org/wiki/Ansible [2] https://eprint.iacr.org/2013/879.pdf