In the attempt to change consensus rules there is a simple set of choices:

1) hard fork: creates a chain split
2) soft fork: creates a chain split
3) 51% attack: does not create a chain split

The presumption being that one can never assume 100% explicit adoption of any rule change.

A 51% attack can of course fail. It is also possible that signaling can be untruthful. But miner signaling provides some level of assurance that it will be successful. This level of assurance is increased by adoption of a higher than majority threshold, as has been done in the past.

Most of the discussion I’ve seen has been focused on who is in charge. Bitcoin requires no identity; anyone can mine and/or accept bitcoin - nobody is in charge.

The majority of those who mine can choose to enforce censorship any time they want. They don’t need anyone’s permission. No power is given to them by developers or anyone else. They have that power based on their own capital invested.

Similarly, the economy (those who accept bitcoin) can enforce any rule change it wants to. And it can do so at any level of participation that wants to go along. Anyone can do this, it requires nobody’s permission. Furthermore, it is possible for the economy to signal its level of agreement in every transaction, as miners have done in blocks previously.

But if the objective is to produce a rule change while avoiding a chain split, 50% is a much lower bar than 100%. If there is some other objective, it’s not clear to me what it is.

e

On Feb 28, 2021, at 12:02, Jeremy via bitcoin-dev <bitcoin-dev@lists.linuxfoundation.org> wrote:


Miners still can generate invalid blocks as a result of SPV mining, and it could be profitable to do "bad block enhanced selfish mining" to take advantage of it.


Hard to analyze exactly what that looks like, but...

E.g., suppose 20% is un-upgraded and 80% is upgraded. Taking 25% hashrate to mine bad blocks would mean 1/4th of the time you could make 20% of the hashrate mine bad blocks, overall a > 5% (series expansion) benefit. One could analyze out that the lost hash rate for bad blocks only matters for the first difficulty adjustment period you're doing this for too, as the hashrate drop will be accounted for -- but then a miner can switch back to mining valid chain, giving themselves a larger % of hashrate.

So it is still possible that an un-upgraded miner will fail part 3, and attempting to accommodate un-upgraded miners leads to some nasty oscillating hashrate being optimal.




On Sun, Feb 28, 2021 at 11:52 AM Matt Corallo <lf-lists@mattcorallo.com> wrote:
Note further that mandatory signaling isn't "just" a flag day - unlike a Taproot flag day (where miners running Bitcoin
Core unmodified today will not generate invalid blocks), a mandatory signaling flag day blatantly ignores goal (3) from
my original post - it results in any miner who has not taken active action (and ensured every part of their often-large
infrastructure has been correctly reconfigured) generating invalid blocks.

As for "Taproot" took too long, hey, at least if its locked in people can just build things assuming it exists. Some
already are, but once its clearly locked in, there's no reason to not continue other work at the same time.

Matt

On 2/28/21 14:43, Jeremy via bitcoin-dev wrote:
> I agree with much of the logic presented by Matt here.
>
> BIP8 was intended to be simpler to agree on to maintain consensus, yet we find ourselves in a situation where a "tiny"
> parameter has the potential to cause great network disruption and confusion (rationality is not too useful a concept
> here given differing levels of sophistication and information). It is therefore much simpler and more likely to be
> universally understood by all network participants to just have a flag day. It is easier to communicate what users
> should do and when.
>
> This is ultimately not coercive to users because the upgrade for Taproot itself is provable and analyzable on its own,
> but activation parameters based on what % of economically relevant nodes are running an upgrade by a certain date are
> not. Selecting these sorts of complicated consensus parameters may ultimately present more opportunity for a cooptable
> consensus process than something more straightforward.
>
>
> That said, a few points strike me as worth delving into.
>
>
> 1) Con: Mandatory signalling is no different than a flag day. Mandatory signaling is effectively 2 flag days -- one for
> the signaling rule, 1 for the taproot type. The reason for the 2 week gap between flag day for signaling and flag day
> for taproot rules is, more or less, so that nodes who aren't taproot ready at the 1st flag day do not end up SPV mining
> (using standardness rules in mempool prevents them from mining an invalid block on top of a valid tip, but does not
> ensure the tip is valid).
> 2) Con: Releasing a flag day without releasing the LOT=true code leading up to that flag day means that clients would
> not be fully compatible with an early activation that could be proposed before the flag day is reached. E.g., LOT=true
> is a flag day that retains the possibility of being compatible with other BIP8 releases without changing software.
> 3) Pro: BIP-8 is partially in service of "early activation" and . I'm personally skeptical that early activation is/was
> ever a good idea. A fixed activation date may be largely superior for business purposes, software engineering schedules,
> etc. I think even with signaling BIP8, it would be possibly superior to activate rules at a fixed date (or a quantized
> set of fixed dates, e.g. guaranteeing at least 3 months but maybe more).
> 4) Pro: part of the argument for BIP-8=false is that it is possible that the rule could not activate, if signaling does
> not occur, providing additional stopgap against dev collusion and bugs. But BIP-8 can activate immediately (with start
> times being proposed 1 month after release?) so we don't have certainty around how much time there is for that secondary
> review process (read -- I think it isn't that valuable) and if there *is* a deadly bug discovered, we might want to
> hard-fork to fix it even if it isn't yet signaled for (e.g., if the rule activates it enables more mining reward). So I
> think that it's a healthier mindset to release a with definite deadline and not rule out having to do a hard fork if
> there is a grave issue (we shouldn't ever release a SF if we think this is at all likely, mind you).
> 5) Con: It's already taken so long for taproot, the schedule around taproot was based on the idea it could early
> activate, 2022 is now too far away. I don't know how to defray this other than, if your preferred idea is 1 year flag
> day, to do that via LOT=true so that taproot can still have early activation if desired.
>
> Overall I agree with the point that all the contention around LOT, makes a flag day look not so bad. And something
> closer to a flag day might not be so bad either for future forks as well.
>
> However, I think given the appetite for early activation, if a flag day is desired I think LOT=true is the best option
> at this time as it allows our flag day to remain compatible with such an early activation.
>
> I think we can also clearly communicate that LOT=true for Taproot is not a precedent setting occurence for any future
> forks (hold me accountable to not using this as precedent this should I ever advocate for a SF with similar release
> parameters).
>
>
> _______________________________________________
> bitcoin-dev mailing list
> bitcoin-dev@lists.linuxfoundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>
_______________________________________________
bitcoin-dev mailing list
bitcoin-dev@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev