Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
From: vjudeu@gazeta.pl
To: Erik Aronesty <erik@q32.com>
In-Reply-To: <CAJowKgLx796AWaZAMOGVWFsT2LK8zvCvxiY-sv110gv1cewCwg@mail.gmail.com>
Date: Thu, 11 May 2023 13:02:57 +0200
Message-Id: <171698970-6184d5f773dee0589bf3373c44b9f21f@pmq2v.m5r2.onet>
Cc: Bitcoin Protocol Discussion <bitcoin-dev@lists.linuxfoundation.org>
Subject: Re: [bitcoin-dev] tx max fee
Precedence: list

> confused.   the rule was "cannot pay a fee > sum of outputs with consider=
ation of cpfp in the mempool"
> your example is of someone paying a fee "< sum"  which wouldn't be blocked

Every transaction paying "fee > sum" can be replaced by N transactions payi=
ng "fee <=3D sum", where the sum of all fees will be the same. That means, =
someone will still do the same thing, but it will be just expanded into N t=
ransactions, so you will reach the same outcome, but splitted into more tra=
nsactions. That means, mempool will be even more congested, because for exa=
mple instead of 1kB transaction with huge fee, you will see 100 such transa=
ctions with smaller fees, that will add to the same amount, but will just c=
onsume more space.

> show me how someone could move 1 btc and pay 2 btc as fees...

In the previous example, I explained how someone could move 1k sats and pay=
 almost 1 BTC as fees. But again, assuming that you have 3 BTC, and you mov=
e 1 BTC with 2 BTC fee, that will be rejected by your rules if and only if =
that will be done in a single transaction. But hey, the same owner can prep=
are N transactions upfront, and release them all at the same time, Segwit m=
akes it possible without worrying about malleability.

So, instead of:

3 BTC -> 1 BTC

You can see this:

3 BTC -> 2 BTC -> 1 BTC

If that second transaction will not pass CPFP, more outputs could be used:

+--------------------+--------------------+--------------------+
| 3.0 BTC -> 0.5 BTC | 0.5 BTC -> 0.5 BTC | 0.5 BTC -> 0.5 BTC |
|            0.5 BTC | 0.5 BTC    0.5 BTC | 0.5 BTC            |
|            0.5 BTC | 0.5 BTC            +--------------------+
|                    +--------------------+
|            0.5 BTC | 0.5 BTC -> 0.5 BTC |
|            0.5 BTC | 0.5 BTC            |
+--------------------+--------------------+

As you can see, there are four transactions, each paying 0.5 BTC fee, so th=
e total fee is 2 BTC. However, even if you count it as CPFP, you will get 1=
.5 BTC in fees for the third transaction in the chain. Note that more outpu=
ts could be used, or they could be wired a bit differently, and then if you=
 will look at the last transaction, the sum of all fees from 10 or 15 trans=
actions in that chain, could still pass your limits, but the whole tree wil=
l exceed that. If you have 1.5 BTC limit for that 3 BTC, then you could hav=
e 20 separate chains of transactions, each paying 0.1 BTC in fees, and it w=
ill still sum up to 2 BTC.

> the only way around it is to maintain balances and use change addresses. =
  which would force nft and timestamp users to maintain these balances and =
would be a deterrent

Not really, because you can prepare all of those transactions upfront, as t=
he part of your protocol, and release all of them at once. You don't have t=
o maintain all UTXOs in between, you can create the whole transaction tree =
first, sign it, and broadcast everything at once. More than that: if you ha=
ve HD wallet, you only need to store a single key, and generate all address=
es in-between on-the-fly, as needed. Or even use some algorithm to determin=
istically recreate the whole transaction tree.


On 2023-05-10 19:42:49 user Erik Aronesty <erik@q32.com> wrote:
confused.=C2=A0 =C2=A0the rule was "cannot pay a fee > sum of outputs with =
consideration of cpfp in the mempool"


your example is of someone paying a fee "< sum"=C2=A0 which wouldn't be blo=
cked


note: again, i'm not a fan of this, i like the discussion=C2=A0of "bitcoin =
as money only" and using fee as a lever to do that


show me how someone could move 1 btc and pay 2 btc as fees... i think we ca=
n block it at the network or even the consensus layer, and leave anything b=
ut "non-monetary use cases" intact.=C2=A0 =C2=A0the only way around it is t=
o maintain balances and use change addresses.=C2=A0 =C2=A0which would force=
 nft and timestamp users to maintain these balances and would be a deterrent


im am much more in favor of=C2=A0doing=C2=A0something like op_ctv which all=
ows many users to pool fees and essentially "share" a single utxo.
.
=C2=A0=C2=A0


On Wed, May 10, 2023 at 12:19=E2=80=AFPM <vjudeu@gazeta.pl> wrote:

> possible to change tx "max fee"=C2=A0 to output amounts?

Is it possible? Yes. Should we do that? My first thought was "maybe", but a=
fter thinking more about it, I would say "no", here is why:

Starting point: 1 BTC on some output.
Current situation: A single transaction moving 0.99999000 BTC as fees, and =
creating 1000 satoshis as some output (I know, allowed dust values are lowe=
r and depend on address type, but let's say it is 1k sats to make things si=
mpler).

And then, there is a room for other solutions, for example your rule, menti=
oned in other posts, like this one: https://lists.linuxfoundation.org/piper=
mail/bitcoin-dev/2023-May/021626.html

> probably easier just to reject any transaction where the fee is higher th=
an the sum of the outputs

Possible situation after introducing your proposal, step-by-step:

1) Someone wants to move 1 BTC, and someone wants to pay 0.99999000 BTC as =
fees. Assuming your rules are on consensus level, the first transaction cre=
ates 0.5 BTC output and 0.5 BTC fee.
2) That person still wants to move 0.5 remaining BTC, and still is willing =
to pay 0.49999000 BTC as fees. Guess what will happen: you will see another=
 transaction, creating 0.25 BTC output, and paying 0.25 BTC fee.
...
N) Your proposal replaced one transaction, consuming maybe one kilobyte, wi=
th a lot of transactions, doing exactly the same, but where fees are distri=
buted between many transactions.

Before thinking about improving that system, consider one simple thing: is =
it possible to avoid "max fee rule", no matter in what way it will be defin=
ed? Because as shown above, the answer seems to be "yes", because you can a=
lways replace a single transaction moving 1 BTC as fees with multiple trans=
actions, each paying one satoshi per virtual byte, and then instead of cons=
uming around one kilobyte, it would consume around 1 MvB per 0.01 BTC, so 1=
00 MvB per 1 BTC mentioned in the example above.


On 2023-05-08 13:55:18 user Erik Aronesty via bitcoin-dev <bitcoin-dev@list=
s.linuxfoundation.org> wrote:
possible to change tx "max fee"=C2=A0 to output amounts?


seems like the only use case that would support such a tx is spam/dos type =
stuff that satoshi warned about


its not a fix for everything, but it seems could help a bit with certain at=
tacks=C2=A0