Tari Protocol Discussion 42
On Monday, the Tari community discussed UTXO commitment. Below is the TL;DR on Monday’s conversation (full transcript included below):
- We only require a TXO, but a UTXO can be beneficial
- A MMR looks to be the best compromise between the ways to commit a UTXO
Join us for our next discussion on Freenode in #tari-dev.
Discussion times proposed by the Tari community:
Mondays: 6pm CAT (12pm EST)
Thursdays: 11am CAT (5am EST)
To keep up with the latest Tari developments, you can follow the project on Twitter.
Transcript of Monday discussion
14:05 <Blackwolfsa> Hi Everyone, I thought perhaps we can talk a bit about the need to commit to the UTXO set for later's discussion at 16:00 UTC 14:06 <Blackwolfsa> If you want a bit more background info on this, these are good reads to quickly get some good info: https://sanket1729.github.io/utxo-commitments A survey on UTXO commitments schemes for bitcoin The Unspent Transaction Output (UTXO) set is the subset of Bitcoin transaction outputs that have not been spent at a given moment. Whenever a new transaction... 14:06 <Blackwolfsa> https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2016-May/012715.html [bitcoin-dev] Making UTXO Set Growth Irrelevant With Low-Latency Delayed TXO Commitments 14:08 <Blackwolfsa> Basic problem we need to discuss is how do we commit to the UTXO set, we have a few options: naive hash, merkle tree, MMR, rolling utxo hash, or a mix of the previously mentioned. All have advantages and disatvantages 14:10 <Blackwolfsa> A big constraint we have on it, is that we need to be easily able to rewind it, so we can handle reorgs 16:19 <tar1b0t> [mattermost] <Cayle> We also need to efficiently implements a `is_in_utxo_set(this_utxo)`, i.e. a proof that a given utxo is in the current set 16:19 <tar1b0t> [mattermost] <Cayle> And of course, provide a commitment to the UTXO set 16:27 <@CjS77> Side note: If I can get some agreement on this small PR, that'd be 💯- https://github.com/tari-project/tari/pull/510 Remove RFC-315 by CjS77 · Pull Request #510 · tari-project/tari Description Transfers are a privilege, not a right ;) There is no requirement that assets be transferable, therefore this shouldn't be a generic RFC, but soemthing that is left to the individua... 18:02 <Blackwolfsa> I forgot about prooving a utxo in a spesific set 18:03 <Blackwolfsa> but then we can leave out naive hashing, I dont think thats possible 18:04 <@CjS77> Yeah, it's the worst from a performance pov 18:04 <Blackwolfsa> if want to proove a utxo I think the only real contenders is merkle tree and MMR 18:04 <@CjS77> I've been grokking Grin's code a bit more and it actually looks like they commit to the entire TXO set, not just UTXOs 18:05 <@CjS77> I agree Blackwolfsa 18:05 <Blackwolfsa> I read around that they where looking at commiting to the stxo as well 18:05 <simian_za> Why commit to those? 18:06 <Blackwolfsa> but a MMR well indirectly commit to the whole txo 18:06 <Blackwolfsa> I could not really pickup why 18:06 <@CjS77> ^ this, so that's why I think grin does it 18:07 <@CjS77> It's easier? 🤷♂️ 18:07 <simian_za> So you mean they don't ever discard the data in the leaves of the MMR even if a whole subtree is spent? 18:07 <Blackwolfsa> no you can 18:08 <Blackwolfsa> but only if the whole sub tree is spent can you remove 18:08 <Blackwolfsa> and only the leaves then 18:08 <simian_za> depends on the height of the subtree surely? 18:09 <neonknight64> If one utxo is abandoned then the sub tree can never be removed? 18:09 <Blackwolfsa> yes 18:09 <simian_za> but ok I see that you are commiting to the spent TXOs in that case essentially 18:10 <Blackwolfsa> if you spent the two leaves, you can prune the two leaves and leave the root 18:10 <Blackwolfsa> till you spent the sub tree next to it etc 18:14 <@CjS77> So my question is (and maybe I'm answering it as I write, but anyway).. if we are committing to the whole TXO set (not just UTXOs), is this even safe? (ignoring info leaking concerns).. 18:14 <neonknight64> It would be nice if people can be encouraged to coin join dust utxos rather than abandoning them. 18:15 <@CjS77> e.g. Set at t0 = 1 2 3 4 5 => Commit a1b2... 18:15 <Blackwolfsa> neonknight64 grin was discussing some insentive to encourage it 18:16 <Blackwolfsa> CjS77 I am missing why it is unsafe? I think its the safest we can be 18:16 <@CjS77> Now if 2 is spent, we have 1 x 3 4 5 6 => H(a1b2..| H(6)) 18:17 <@CjS77> Ok, let me say it this way. Changing a UTXO to a STXO doesn't actually change the local peaks hash, correct? 18:18 <@CjS77> i.e. Hash(1 2 3 4 5) = Hash(1 x x 4 5) 18:18 <@CjS77> where x is a STXO 18:19 <Blackwolfsa> yes 18:19 <@CjS77> Our protection from malleability attacks is that you always need a NEW utxo to prevent inflation, so the root hash still changes 18:19 <Blackwolfsa> correct 18:19 <@CjS77> Ok, so if we had zero-valued commitments, you could potentially play some games 18:20 <@CjS77> That's thebasic core point I need to be convinced about before agreeing that MMRs are the way to go 18:20 <Blackwolfsa> well if you add a zero-value commitment you still need to add it the utxo set 18:21 <Blackwolfsa> and you cant just delete a utxo so you will spend it to something if not just a zero value commitment 18:22 <@CjS77> I was thinking when the zero-valued UTXO gets spent; but I get you always have the Coinbase UTXO in a block anyway 18:22 <@CjS77> *guess 18:23 <Blackwolfsa> yes, there is no way where you can just spend any utxo without adding one as well 18:23 <Blackwolfsa> MMR's are add only by design 18:23 <Hansie> Which games cjs77? 18:24 <Blackwolfsa> but because we spend some leaves over time, we can delete some old data since they will never get asked for anymore 18:24 <@CjS77> Ok, now what are the downsides of a straight MMR or sparse MMR? 18:25 <@CjS77> I beg your pardon. Strsight Merkle Tree or Sparse Merkle tree 18:26 <Hansie> Blackwolfsa so the pruning happens randomly as it goes along, but more with older UTXOs? 18:27 <Blackwolfsa> pruning happens with spent utxo's thats random 18:27 <Blackwolfsa> CjS77 I dont see how we can effectivly use sparse tree's 18:27 <Blackwolfsa> and comparing a straight merkle tree to a MMR, the MMR is the the way better one 18:28 <@CjS77> well besides the fact that you must keep hashes of all our TXOs 18:28 <Blackwolfsa> Yes but you dont have to hash the whole utxo every time 18:28 <Blackwolfsa> with a merkle tree you have to 18:30 <@CjS77> For completeness, here's a decent write-up of sparse merkle trees, https://medium.com/@kelvinfichter/whats-a-sparse-merkle-tree-acda70aeb837 What’s a Sparse Merkle Tree? - Kelvin Fichter - Medium If you’ve been around the Ethereum research community lately, you might’ve heard of something called a sparse Merkle tree. They might sound… 18:30 <@CjS77> and I tend to agree that this isn't the right use case for them 18:31 <Blackwolfsa> comparing hashes, at this moment, grin has 5762 utxo's, the last block had 5. You on a worst case senario, you are hashing 25 hashes, compared to 5762 hashes 18:31 <Hansie> They talk about `Proving Non-inclusion` - do we need that? 18:32 <@CjS77> So Blackwolfsa, if we commit to the TXO, and not the UTXO, how do you validate that your UTXO set is correct when you're syncing -- is it indirect? 18:32 <Blackwolfsa> compare merkle roots 18:33 <@CjS77> You know that the UTXO set that you have is _at least_ a subset of the full UTXO set, 18:33 <Blackwolfsa> we commit to that every block 18:33 <@CjS77> That includes Spent and Unspent 18:33 <@CjS77> You can't calculate the root from just a UTXO set, you need the merkle proof as well, no? 18:34 <Blackwolfsa> they are the same 18:34 <Blackwolfsa> the proof of the MMR/tree is the merkle root, and the utxo proof is that as well 18:35 <@CjS77> Hear me out -- the MMR looks like (1 x x 4 5) with root b1a1 18:35 <@CjS77> Now I'm syncing with you, so you send me (1 4 5). 18:35 <Blackwolfsa> no 18:36 <Blackwolfsa> I will have to send you (1, null, null, 4, 5) 18:36 <@CjS77> Ok, sure. But also 1, null's parent and null,4's parent surely 18:36 <Blackwolfsa> or something like (1,1; 4,3;5,5) 18:37 <@CjS77> Cool, that's what I was calling amerkle proof 18:37 <Blackwolfsa> with this tree spesifically you will actually send 1,2,3,4,5 18:37 <@CjS77> You'll send the STXOs? 18:38 <Blackwolfsa> because 1 and 4 was not spent you cannot prune 2,3 yet 18:38 <@CjS77> Do you send the whole thing, or just heir hashes? 18:38 <Blackwolfsa> just the hashes 18:39 <Blackwolfsa> you dont need to send the actual utxo 18:39 <@CjS77> stxo 18:39 <Blackwolfsa> yes 18:40 <@CjS77> right, so you could lie to me and *say* that 3 was spent when it wasn't; just send me the hash and I'd be non-the-wisert 18:40 <@CjS77> The hash would be the same 18:40 <@CjS77> The only time I'd eventually figure out something was wrong was when I summed the UTXOs and checked the accumulated excess 18:41 <Blackwolfsa> potentially 18:41 <Blackwolfsa> but you have to check that anyway I think 18:41 <@CjS77> Yes -- I was being a bit Socratic, but wanted to follow the logic through 18:42 <@CjS77> The point is that you can't actually trust that the UTXO set is complete from the commitment and a provided set when using an MMR 18:42 <@CjS77> you have to do the inflation check too, and then you're ok 18:43 <Blackwolfsa> the mmr proof of the utxo will only help you syncing a block or so, I think it will help you just syncing the whole chain 18:44 <Hansie> Interesting conclusion, that the Merkle proof does not proof the UTXO set. 18:44 <@CjS77> When you sync, you get the UTXO set at a given block, don't you? 18:44 <Blackwolfsa> yes 18:45 <@CjS77> Hansie, only because it's a commitment to the whole TXO set, not just UTXOs 18:45 <Blackwolfsa> But that might be why grin was investiagting commiting the stxo 18:46 <@CjS77> I've got to run, but it would be great to continue this line of thinking. 18:46 <Blackwolfsa> https://github.com/mimblewimble/grin/issues/859 Question - STXO MMR (aka the rm_log)? · Issue #859 · mimblewimble/grin [Just thinking out loud here, but wanted to put it somewhere for reference.] Right now we have a "remove log" and a "prune list" but we do not explicitly commit to these anywher... 18:46 <Hansie> Makes sense, thanks for the interesting discussion 18:46 <Blackwolfsa> thats them discussing the issue, they have some solution to it, but I have not yet completly dug through how it works 18:46 <@CjS77> ^^ Nice, thanks