Gold collapsing. Bitcoin UP.

cypherdoc

Well-Known Member
Aug 26, 2015
5,257
12,995
@Mengerian

One aspect to the utxo that I've been kicking around in my head for a while now is the relationship between tx fees, bitcoin valuation, and set size. It seems like alot of perverse things could happen under a forced fee market wrt to maintaining utxo bloat by essentially stranding outputs as economically infeasible to spend. Massive waves of price appreciation (and concurrent drops in tx fees in nominal terms) seem like a natural mechanism to counter this, assuming tx volumes are not intentionally pegged at capacity. This is something Mike Hearn has gotten heavily attacked for well after the fact, but cutting the default min relay fee after the fall 2013 rally probably had really positive effects.

For example, I bet that there are quite a few small hobbyist pool miners from late 2013 / early 2014 who maybe played around with very modest setups of a few GH/s with Block Erupters and such who weren't very saavy about it, somewhere got the impression that they should be paranoid about leaving any balance at all on a pool, so they have a tremendous number of tiny outputs to the same address. I have a friend who's exactly in that boat, sitting on a bunch of 0.01 outputs from BTCGuild.
@Mengerian @albin

this is a great discussion.

i think it helps to go back and review the history of the UTXO set. iirc, i'm almost certain this was pwuille's innovation. at the time (& probably still) it made sense to bring the unspent tx outputs into fast access RAM & dbcache so as to validate incoming tx's faster & more efficiently. not sure exactly how it was done before but maybe your node had to go back to the blockchain itself on disk & do a lookback to the actual block containing the txid of the output referenced which undoubtedly takes more time. so if you think about it, the UTXO is just a subset of the blockchain. it happens to be the most important part of the blockchain that needs fast access & is contained in the chainstate folder of your disk. so as the blockchain grows from increased adoption, it's only natural for the UTXO set to grow with it. and if more ppl use Bitcoin, then naturally they will inherit smaller and smaller numbers of satoshi's for their individual usages years into the future as the bitcoin units propagate (spread) out to the global population. and yes, it's only natural (mandatory in fact) in that scenario that those individual smaller satoshi's will have to be more valuable, in fiat terms, for them to be used efficiently and with reasonable tx fees. otherwise, they get stuck as economically unviable.

so to make an attempt, as with SWSF, to truncate the growth of this set by giving cost discounts to tx's that consolidate the UTXO's is tantamount to crippling Bitcoin's growth. it would be like acknowledging that a baby's head has to grow in the same proportion to the rest of it's body as they get older but deciding somewhere along the line that that's not a good idea b/c your head is the heaviest part of the human body. "let's consolidate all your brain cells b/c we don't want you to get to adult size and have to carry around all that weight". it makes no sense. unless of course, your goal is to divert those brain cells to a second head called LN.

on to another pt. i've heard this "SW makes inputs cheaper & outputs more expensive" from several ppl and it's confused me b/c to me the argument doesn't make sense. fact: these "cost discounts" are being given to those tx's which create larger signatures per tx. those are multisigs. what do multisigs have to do with consolidating inputs (UTXO's)? that's why i asked @Mengerian that question and his answer is that they don't do that necessarily. and i agree. so it appears to me that this may just be a red herring to convince miners that this theoretical UTXO reduction advantage is for them since they can point to f2pool's 1MB 25 sec tx of several months ago and say, "hey, you might want to do some of these too!". so then i started poking around about just how frequently these types of multi-input tx's get done by miners, exchanges. it appears to me, not much. i asked jtoomim the other day. here's his response, emphasis mine:

cypherdoc 3:59 PM] how often do pools like f2pool do those multi input complex sig tx's? didn't i hear you say maybe just once a quarter or even once a year?

jtoomim4:17 AM] multi-input huge txs? that's only been done once, as a spam cleanup

multi-output huge txs? once a year. multi-output big txs? once a day.

doesn't sound like often to me. so then, if this is the case, what is this discount for? like i've been saying, it's to advantage LN multisigs and this drive to create a offchain tx's LN that might be cheaper. so here we have evidence of another inconsistency in theory and logic from Blockstream core dev. diversion-->cheaper inputs & more expensive outputs (reduce UTXO's!!!). reality-->advantage LN multisigs.

one further pt. to the extent that UTXO set growth is a "problem". shouldn't we be blaming pwuille for this? i'm being snarky with this b/c small blockists are quick to blame Hearn for the levelDB fork of March 2013, as if he was the only one who signed off on it. what has happened in reality is that levelDB was quickly fixed and still adopted and is what we use to this day to our benefit. similarly, the UTXO set was probably a good idea and as @Mengerian says, it's growth is a good thing and a reflection of growth in the Bitcoin economy. there have been many suggestions over the years as to how to deal with this growth that we should explore going forward; commitments for one. right now, it's just not a problem.
 
Last edited:

cypherdoc

Well-Known Member
Aug 26, 2015
5,257
12,995
@sgbett

which reminds me.

@bitsko dropped this tweet bomb of mine just the other day directly into the middle of BitcoinCore Slack General chat section in which Patrick Strateman was online and active & the only responses it generated were:

1. wall of text
2. my eyes hurt-Strateman
3. cypherdoc *!@%$#
4. not worth reading if he can't prove his relevance

note: you'd think they'd be interested in the math & economic consequences but i think it's surprisingly & painfully evident, even from the guys on their side, how little they understand of both and reflects their willingness to throw their support behind gods like pwuille and gmax w/o question:

 

sgbett

Active Member
Aug 25, 2015
216
786
UK
I think I've been accused of having cognitive dissonance in the past, and I've come to accept it's true.

All the evidence points to shenanigans, but I always (ok not always, but mostly try to) recoil from accusing people of being shills, or having ulterior motives, or any of the other shitty behaviour because I feel that doing so would be descending to that level so I'm trying hard to stay on topic. I really want to believe that people are inherently good and its all just some big misunderstanding. The 1.5MB suggestion though might has well have had /trollface as the synopsis. :/
 
  • Like
Reactions: majamalu and solex

sickpig

Active Member
Aug 28, 2015
926
2,541
@cypherdoc

from the nitpick dept: technically 2013 HF wasn't due to LevelDB but to BerkeleyDB. Actually to the default bitcoin core default BerkeleyDB settings.

In fact the nodes belonging to 0.7 fork rejected a perfectly valid block due to the BerkeleyDB inability to store it.

if every 0.7 full node had followed this:

https://bitcointalk.org/index.php?topic=152208.msg1615346#msg1615346

the downgrade of the 0.8 fork wouldn't have been needed
 

Justus Ranvier

Active Member
Aug 28, 2015
875
3,746
@sickpig
You're confusing two things. Pool operators and solo miners are only a _tiny_ subset of the users of the system, ones whos motivations are not always well aligned with everyone elses and are often only aligned by all the things they cannot control. For example, if it were really just up to miners— 50 BTC _forever_. But fortunately no one elses nodes would accept their inflationary blocks. Amusingly, turning things over to a majority of core developers would make it depend on more people than a majority of hashpower (which is very close to being decided by a single person right now); not that either is remotely acceptable.

Bitcoin is a participatory system which ought to respect the right of self determinism of all of its users.
What a difference a few years makes.
 

cypherdoc

Well-Known Member
Aug 26, 2015
5,257
12,995
the double short ETN on oil is going nuts. that's very bad news:

 

Mengerian

Moderator
Staff member
Aug 29, 2015
536
2,597
there have been many suggestions over the years as to how to deal with this growth that we should explore going forward; commitments for one. right now, it's just not a problem.
Yeah, I agree with this. It is not a problem yet, so something like the SW consensus rule formula is a premature optimization. Whether the "discounting factor" is 1/4 or 1/2 is pretty much irrelevant right now. Imposing a new consensus rule to address this non-problem is the wrong approach, we should be moving towards managing these resources more intelligently.

@albin: Your example of your friend's dust transaction miner payouts shows that there are already incentives in place discouraging dust outputs. Miners can set policy to discourage these types of transaction, and we can see it happening already. It is possible to imagine many ways that nodes and miners could implement policies to deal with UTXO "bloat". Maybe in future they won't keep the whole UTXO set in RAM, saving lower-priority outputs to disk. Transactions spending theses outputs would then take longer to verify, and require higher fees.

It just seems like there are so many ways to deal with UTXO growth that can evolve organically over time. We don't need to centrally engineer this issue. Solutions and improvements can arise in a decentralized manner from individual incentives, node relay behavior, and miner policy.
 

cypherdoc

Well-Known Member
Aug 26, 2015
5,257
12,995
users can help consolidate UTXO's by themselves by just creating a multi input tx. what's key is to keep tx fees low so that it doesn't become too expensively prohibitive to do this. the 1MB fee mkt forcing higher fees & suppressing exchange value destroys this whole ability and can strand low value UTXO's.
 
Last edited:

solex

Moderator
Staff member
Aug 22, 2015
1,558
4,693
Yes, the UTXO must increase with ecosystem growth and diversification of holdings, but ideally the increase is not simply "spam".

The fix for this problem was an early change by Lee after creating Litecoin:
http://cointelegraph.com/news/litecoin-shows-there-is-a-simple-fix-for-spam-attacks-on-bitcoin
The fix implemented in Litecoin is just to charge the sender a fee for each tiny output he creates. For example, in this specific attack, the sender is charged one fee for sending to 34 tiny outputs of 0.00001 BTC. With the fix, that fee would be 34 times as much. So it would cost the attacker a lot more to perform the spam attack. The concept is fairly simple: the sender should pay for each tiny output he/she creates.
Does anyone have a simple reason why this is not a good idea? Has the Bitcoin UTXO become much more bloated than otherwise, i.e. if this change was done?
 

sickpig

Active Member
Aug 28, 2015
926
2,541
@Justus Ranvier

yours is even better.

I wonder if this this gmax quotes are good enough to justify a post on r/btc...
 
  • Like
Reactions: AdrianX

sickpig

Active Member
Aug 28, 2015
926
2,541
Actual network hash rate estimate is ~1,195,996,406 GH/s

On December 5 2015 the estimated hash rate was ~ 566,236,898 GH/s

It doubled
in just two months.

I was thinking that recent jump in difficulty was due to Bitfury new deployment, but to my surprise it seems that the mass production of their new chips is not complete, hence they are not selling yet.

If this is true we could reach 2 EH/s by March/April 2016 easy (E = exa = 1k peta).

if I have to be honest I'm amazed.
 

cypherdoc

Well-Known Member
Aug 26, 2015
5,257
12,995
i just wanna say, we can anticipate that there won't be a decentralized LN network. namely, b/c there aren't any obvious solutions to the decentralized routing problem at present and the fact that to make money, there will have to be LN hubs that get in the middle of every tx to make fees. this is competition to miners. that's not good.

we can use the example of what appears to be happening to SC's. instead of the spvp model via a 2wp, we instead see Liquid, a centralized federated server hub which has invited select permissioned Bitcoin exchanges to participate. too bad if you weren't invited. Liquid charges fees to it's clients. there's also this push for SW so that Blockstream can insert Confidential Transactions directly into the protocol thru a SF script versioning trick. point being, the spvp 2wp model doesn't look promising. in that light, alot is riding on LN for Blockstream:

 

Bloomie

Administrator
Staff member
Aug 19, 2015
510
803
Live from BitDevs NYC. Didn't take long for allegations of politicking to start flying. :)


[doublepost=1455066203,1455065306][/doublepost]@Peter R