Re: [bitcoin-dev] Towards a means of measuring user support for Soft Forks

From: Keagan McClelland <keagan.mcclelland@gmail•com>
To: micaroni@gmail•com
Cc: Bitcoin Protocol Discussion <bitcoin-dev@lists•linuxfoundation.org>
Subject: Re: [bitcoin-dev] Towards a means of measuring user support for Soft Forks
Date: Wed, 27 Apr 2022 12:32:33 -0600	[thread overview]
Message-ID: <CALeFGL3oV2c8cioak1-CfDxNp93wz1qwasYETTgkAY+C31Mbxw@mail.gmail.com> (raw)
In-Reply-To: <CAHvMVPQWkEN_tpwNwkVkPs9W=DsmkBxMfJfqAbv8JNPVxkfKkQ@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 14840 bytes --]

Felipe,

> For me, the consensus should follow the current line: discussions and
tests carried out by experts. We all know that the most important devs have
the most weight in discussions. And that's how it should be, because they
understand far better than any other lowly mortal. Consensus simply means
that there are not at least two or three important people opposing the idea
with solid arguments. Is it very subjective and difficult? Yes. For sure.
We all yearn for objective answers or methods. However, any method would
fail. At the end, after numerous discussions and an apparent consensus, the
objective answer and the real consensus will be obtained in the network, in
the nodes upgrading. If there is a big war, the network will end up
splitting in two, as it has in the past. To avoid any unwanted splits we
discuss for exhaustion here in the list.

This is essentially an admission that devs have control over the protocol.
Users "having control" but deferring their judgement to devs is not
meaningfully different than devs "having control". Many people have
asserted, quite strongly, that this ought not be how Bitcoin governs
itself. I myself am on the fence about what is practically possible or not.
However, let's say that your supposition is correct. How would we protect
against a corollary scenario where a dev has a proposal that looks great
but has dark ends that no one notices yet, if the process for evaluation
more or less is to defer to "the most important devs" expertise? Presumably
we hash this out in forums like this, but in order to "override" the "most
important devs" we have to have a way (formalized or not) of deciding when
the "lesser experts" in aggregate have better judgement.

Erik,

> There are many challenges with on-chain voting, here are a few:

This may be hair-splitting but I feel it important to clarify that my
proposal isn't voting per se. Calling it that doesn't bug me, but the
mechanics are meaningfully different than a simple tally vote which is the
intuition that I think that term conveys. As Billy mentions this proposal
actually requires that miners block signals from inclusion in the block if
they themselves do not signal. I'm not necessarily claiming this is a
superior design overall, however the "flaw" you point out is by design in
this case. My goal in the proposal was really to give users a means of
applying direct economic pressure to miners, who do inevitably play a role
in BIP8/BIP9 activation procedure.

Ryan,

> - you're feeding the Chainalysis beasts, when hodlers move their UTXOs;

Definitely a frightening proposition I hadn't considered. It does open up
the possibility of tracking individual preferences and targeting of
political opponents.

>   - yuk, it's voting.

I don't think the process of collecting information on user preference is
in and of itself bad. Where I think Bitcoiners really want to avoid voting
is this notion that 51% of the constituency can bully the other 49% into
whatever they want. No part of my proposal suggests this, nor is it
something I would want.

-----

I think there are a few questions surrounding the issue of soft fork
activation. Perhaps it warrants zooming out beyond even what my proposal
aims to solve. In my mind the most important questions surrounding this
process are:

1. In an ideal world, assuming we could, with perfect certainty, know
anything we wanted about the preferences of the user base, what would be
the threshold for saying "this consensus change is ready for activation"?
    1a. Does that threshold change based on the nature of the consensus
change (new script type/opcode vs. block size reduction vs. blacklisting
UTXOs)?
    1b. Do different constituencies (end users, wallets, exchanges,
coinjoin coordinators, layer2 protocols, miners) have a desired minimum or
maximum representation in this "threshold"?
2. Given an answer from #1, what tests can we devise to measure those
levels of support directly? If we can't measure it directly, can we measure
different indicators that would help us infer or solve for the knowledge we
want?
3. Can any of the answers to #2 be "gamed"? I'm defining "game" here to
mean that the measurement taken, diverges from the ground truth we are
trying to get at in such a way that its divergence would be undetectable.

If we do not answer these sorts of questions we can get technical consensus
through this messy process, but when it comes to assessing user consensus,
it is just going to devolve into dogma and demagoguery as we each have our
own perceptions or agendas and there is no rigorous way for anyone to
refute our claims. This would, again, be an admission that devs ultimately
do make protocol decisions. Perhaps it's unavoidable and we are doomed to
this painful process of arguing with one another until there's only one
opinion left standing (either because of merit or just plain old grit).
However, if this is the case, I don't think we can honestly claim that devs
don't control the protocol (as a group).

I don't think we will have broad agreement on #1 as it is ultimately a
value judgement and even the most intellectually honest people in Bitcoin
dev are going to have different value sets. I think this is OK, to a
degree. But where a lot of communication breakdown occurs is when people
are debating the properties of #2/#3 when they don't even know that there
is disagreement between them on #1. I think that everyone having an
individual answer to #1 can make these discussions go a lot more smoothly
in the technical sphere since I think most people can suspend their own
values for the sake of analyzing the effectiveness of a particular
approach. I am concerned, however, that if value differences are allowed to
be passed off as technical evaluations, the quality of the conversation may
erode to the point where no meaningful advancement can happen anymore,
since we will lose our shared framework for understanding. If this occurs
too soon, I believe quite strongly that Bitcoin will be captured through
the increasing power of custodial institutions.

Keagan

On Wed, Apr 27, 2022 at 11:22 AM <micaroni@gmail•com> wrote:

> The idea seems interesting at first glance, but soon we see several
> problems. The biggest problem with votes of this type is that they can be
> easily manipulated. Imagine a powerful attacker who impersonates someone in
> good faith and arrives with a proposal that looks great but has dark ends
> behind it (and that no one has simply noticed yet). It would be enough for
> this attacker to convince major wallets, major exchanges and even
> individuals to believe him. It could be with a good marketing campaign or
> even buying these people. This would create a "false consensus", a
> misconception of what consensus means.
>
> For me, the consensus should follow the current line: discussions and
> tests carried out by experts. We all know that the most important devs have
> the most weight in discussions. And that's how it should be, because they
> understand far better than any other lowly mortal. Consensus simply means
> that there are not at least two or three important people opposing the idea
> with solid arguments. Is it very subjective and difficult? Yes. For sure.
> We all yearn for objective answers or methods. However, any method would
> fail. At the end, after numerous discussions and an apparent consensus, the
> objective answer and the real consensus will be obtained in the network, in
> the nodes upgrading. If there is a big war, the network will end up
> splitting in two, as it has in the past. To avoid any unwanted splits we
> discuss for exhaustion here in the list.
>
> I don't think flagging transactions would be a good method to measure this
> sort of thing. You are handing important technical discussions into the
> hands of those who have no idea about the subject.
>
> Felipe.
>
> On Tue, Apr 26, 2022 at 5:12 PM Keagan McClelland via bitcoin-dev <
> bitcoin-dev@lists•linuxfoundation.org> wrote:
>
>> Hi all,
>>
>> Alongside the debate with CTV right now there's a second debate that was
>> not fully hashed out in the activation of Taproot. There is a lot of
>> argument around what Speedy Trial is or isn't, what BIP8 T/F is or isn't
>> etc. A significant reason for the breakdown in civility around this debate
>> is that because we don't have a means of measuring user support for
>> proposed sof-fork changes, it invariably devolves into people claiming that
>> their circles support/reject a proposal, AND that their circles are more
>> broadly representative of the set of Bitcoin users as a whole.
>>
>> It seems everyone in this forum has at one point or another said "I would
>> support activation of ____ if there was consensus on it, but there isn't".
>> This statement, in order to be true, requires that there exist a set of
>> conditions that would convince you that there is consensus. People have
>> tried to dodge this question by saying "it's obvious", but the reality is
>> that it fundamentally isn't. My bubble has a different "obvious" answer
>> than any of yours.
>>
>> Secondly, due to the trauma of the block size wars, no one wants to utter
>> a statement that could imply that miners have any influence over what
>> rulesets get activated or don't. As such "miner signaling" is consistently
>> devalued as a signal for market demand. I don't think this is reasonable
>> since following the events of '17  miners are aware that they have the
>> strong incentive that they understand market demand. Nevertheless, as it
>> stands right now the only signal we have to work with is miner signaling,
>> which I think is rightly frustrating to a lot of people.
>>
>> So how can we measure User Support for a proposed rule change?
>>
>> I've had this idea floating around in the back of my head for a while,
>> and I'd like to solicit some feedback here. Currently, all forms of
>> activation that are under consideration involve miner signaling in one form
>> or another. What if we could make it such that users could more directly
>> pressure miners to act on their behalf? After all, if miners are but the
>> humble servants of user demands, this should be in alignment with how
>> people want Bitcoin to behave.
>>
>> Currently, the only means users have of influencing miner decisions are
>> A. rejection of blocks that don't follow rules and B. paying fees for
>> transaction inclusion. I suggest we combine these in such a way that
>> transactions themselves can signal for upgrade. I believe (though am not
>> certain) that there are "free" bits in the version field of a transaction
>> that are presently ignored. If we could devise a mapping between some of
>> those free bits, and the signaling bits in the block header, it would be
>> possible to have rules as follows:
>>
>> - A transaction signaling in the affirmative MUST NOT be included in a
>> block that does not signal in the affirmative
>> - A transaction that is NOT signaling MAY be included in a block
>> regardless of that block's signaling vector
>> - (Optional) A transaction signaling in the negative MUST NOT be included
>> in a block that signals in the affirmative
>>
>> Under this set of conditions, a user has the means of sybil-resistant
>> influence over miner decisions. If a miner cannot collect the fees for a
>> transaction without signaling, the user's fee becomes active economic
>> pressure for the miner to signal (or not, if we include some variant of the
>> negative clause). In this environment, miners could have a better view into
>> what users do want, as would the Bitcoin network at large.
>>
>> Some may take issue with the idea that people can pay for the outcome
>> they want and may try to compare a method like this to Proof of Stake, but
>> there are only 3 sybil resistant mechanisms I am aware of, and any "real"
>> view into what social consensus looks like MUST be sybil resistant:
>>
>> - Hashpower
>> - Proof of personhood (KYC)
>> - Capital burn/risk
>>
>> Letting hashpower decide this is the thing that is currently contentious,
>> KYC is dead on arrival both on technical and social grounds, which really
>> just leaves some means of getting capital into the process of consensus
>> measurement. This mechanism I'm proposing is measurable completely
>> en-protocol and doesn't require trust in institutions that fork futures
>> would. Additionally it could be an auxiliary feature of the soft fork
>> deployment scheme chosen making it something you could neatly package all
>> together with the deployment itself.
>>
>> There are many potential tweaks to the design I propose above:
>> 1. Do we include a notion of negative signaling (allowing for the
>> possibility of rejection)
>> 2. Do we make it such that miner signaling must be congruent with >X% of
>> transactions, where congruence is that the signal must match any
>> non-neutral signal of transaction.
>>
>> Some anticipated objections:
>>
>> 1. signaling isn't voting, no deployment should be made without consensus
>> first.
>> - yeah well we can't currently measure consensus right now, so that's not
>> a super helpful thing to say and is breeding ground for abuse in the form
>> of certain people making the unsubstantiated claim that consensus does or
>> does not exist for a particular initiative
>>
>> 2. This is just a proposal for "pay to play", we should not let the
>> wealthy make consensus decisions.
>> - I agree that wealth should not be able to strong-arm decision making.
>> But the status quo seems even worse where we let publicly influential
>> people decide consensus in such a way where not only do they not "lose
>> ammunition" in the process of campaigning, but actually accrue it, creating
>> really bad long-term balances of power.
>>
>> 3. Enforcing this proposal requires its own soft fork.
>> - Yes. It does...and there's a certain cosmic irony to that, but before
>> we consider how to make this happen, I'd like to even discuss whether or
>> not it's a good idea.
>>
>> 4. This gives CoinJoin pool operators and L2 protocol implementations
>> power over deciding consensus.
>> - I see this as an improvement over the status quo
>>
>> 5. This encourages "spam"
>> - If you pay the fees, it's not spam.
>>
>> The biggest question I'd like to pose to the forum is:
>> - Does a scheme like this afford us a better view into consensus than we
>> have today?
>> - Can it be gamed to give us a *worse* view into consensus? How?
>> - Does it measure the right thing? If not, what do you think is the right
>> thing to measure? (assuming we could)
>> - Should I write a BIP spec'ing this out in detail?
>>
>> Cheers,
>> Keagan
>> _______________________________________________
>> bitcoin-dev mailing list
>> bitcoin-dev@lists•linuxfoundation.org
>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>>
>

[-- Attachment #2: Type: text/html, Size: 18374 bytes --]