From: defenwycke <cal.defenwycke@gmail.com>
To: Bitcoin Development Mailing List <bitcoindev@googlegroups.com>
Subject: Re: [bitcoindev] Draft BIP: DustSweep – policy-only UTXO dust compaction
Date: Sat, 13 Dec 2025 06:56:31 -0800 (PST) [thread overview]
Message-ID: <841d6882-9853-4d96-84e8-c4742e17e969n@googlegroups.com> (raw)
In-Reply-To: <02d7368e-95d3-4185-b70f-3aa9b5df1e1d@murch.one>
[-- Attachment #1.1: Type: text/plain, Size: 11112 bytes --]
Hello Murch,
Thanks for the thoughtful response. Please find my response below.
> What does “dust-class” mean? Are you using the Bitcoin Core dust limit or
talking about small amounts in general? I don’t have figures off the top of
my head, but I would assume that there are relatively few UTXOs smaller
than BitcoinCore’s dust limit.
I’m not referring to Bitcoin Core’s relay dust limits here. By “dust-class”
I mean outputs that are economically irrational to spend under typical fee
conditions, even though they remain technically valid. Using current
standard input sizes (≈68 vB for P2WPKH, ≈57.5 vB for P2TR keypath, ≈148 vB
for P2PKH), the break-even spend cost quickly rises into the hundreds of
sats once fee rates exceed a few sat/vB. At more common feerates (≈5–10
sat/vB), outputs below roughly 500–1500 sats are rationally abandoned
depending on script type. So this is an economic classification, not a
relay-policy one.
> You might want to clarify that you mean only P2TR KP inputs. Or would
P2TR SP be permitted?
Yes — only P2TR key-path spends would be permitted. Script-path spends
would be excluded. I’ll make that explicit.
> It would be a lot of work to have a separate pool for this, and I don’t
see a reason why they couldn’t just go in the regular mempool.
Agreed. A physically separate mempool is not required. The intent is simply
that these transactions sit at the very bottom of the normal mempool’s
priority ordering and are treated as lowest priority for eviction and
inclusion
> That said, at 50% full, there are still around ~30 blocks worth of
transactions waiting in the mempool that pay fees, …
Right - and DustSweep is not intended to provide any liveness guarantees in
that situation. These transactions are explicitly opportunistic and are
expected to idle or expire during sustained congestion. That behaviour is
acceptable and consistent with the goal of ensuring they never compete with
fee-paying transactions. The 50% figure was meant as an illustrative policy
threshold, not a claim that blockspace is otherwise unused.
> …the only ones I have seen lately are miners using a minimum feerate of 1
s/vB for their block templates.
That aligns with the intent. DustSweep transactions would only ever be
eligible after normal block assembly, and only in templates that already
include all available fee-paying transactions.
> I assume the intention is to only relay these transactions when there are
blocks that aren’t full, to limit the bandwidth-wasting vector this feature
introduces, but overall it seems to me that it would be most likely for
such transactions to sit in nodes’ memory until they expire.
That’s a fair characterization, and it matches the design goals. To further
limit policy complexity and relay churn, DustSweep transactions would also
be constrained to:
- Confirmed inputs only (no unconfirmed ancestors
- RBF disabled (no replacement or package churn)
- No CPFP assumptions
This keeps them cheap to reason about for node operators, and expiry
without confirmation is an expected outcome rather than a failure mode.
Importantly, these constraints mean DustSweep transactions require no
additional mempool state tracking, package evaluation, or replacement logic
beyond what nodes already implement today.
> It doesn’t seem obvious to me that saving a few dozen sats would greatly
foster the users’ urge to consolidate. It feels like a lot of overhead for
such a small incentive to the users, and relying on the miners to give away
blockspace below market value feels a bit optimistic as well.
I agree that the incentive is not primarily about recovering value.
Empirically, outputs in this range represent a large number of UTXOs but
very little aggregate bitcoin value. Even aggressive consolidation would
recover well under a single BTC in total. The motivation is instead about
long-term UTXO set hygiene: providing a narrow, predictable mechanism for
compacting outputs that are otherwise rationally abandoned, without
displacing market transactions or altering fee dynamics. Because the
economic value involved is small, the mechanism is intentionally
constrained to avoid creating meaningful incentives for either users or
miners to game block construction or relay policy. The benefit is therefore
not measured in recovered bitcoin value, but in avoided long-term UTXO
growth and reduced steady-state resource costs for nodes.
Separately (and not a dependency of this proposal), public analysis of
inscription-related activity shows that a significant share of UTXO growth
is tied to metadata-heavy patterns. Future work on segregated data lanes
could allow voluntary compaction of those UTXOs while preserving metadata,
but that’s orthogonal to DustSweep itself.
Kind regards,
Defenwycke
On Friday, December 12, 2025 at 11:19:17 PM UTC Murch wrote:
> Hey Defenwycke,
>
> > all inputs are “dust-class” UTXOs
>
> What does “dust-class” mean? Are you using the Bitcoin Core dust limit
> or talking about small amounts in general? I don’t have figures off the
> top of my head, but I would assume that there are relatively few UTXOs
> smaller than Bitcoin Core’s dust limit.
>
> > only standard scripts (P2PKH / P2WPKH / P2TR)
>
> You might want to clarify that you mean only P2TR KP inputs. Or would
> P2TR SP be permitted?
>
> > Nodes place these in a small, separate sub-mempool. They’re only
> > accepted when the normal mempool is <50% full, and they’re
> > automatically evicted if normal mempool usage hits 95%.
>
> It would be a lot of work to have a separate pool for this, and I don’t
> see a reason why they couldn’t just go in the regular mempool. If the
> mempool fills up, they’d have the lowest feerates and they’d get kicked
> out first anyway. That said, at 50% full, there are still around ~30
> blocks worth of transactions waiting in the mempool that pay fees, …
>
> > Miners can include them up to a small weight fraction (I suggest ~5%)
> but only after filling the block with regular fee-paying transactions.
>
> … so if they are only considered in blocks that aren’t full, the only
> ones I have seen lately are miners using a minimum feerate of 1 s/vB for
> their block templates. Looking at some popular mempool statistic sites,
> in the past 32 months, there would have only been organically non-full
> blocks between April and August this year.
>
> I assume the intention is to only relay these transactions when there
> are blocks that aren’t full, to limit the bandwidth-wasting vector this
> feature introduces, but overall it seems to me that it would be most
> likely for such transactions to sit in nodes’ memory until they expire.
>
> All that said, at the new minimum feerate of 0.1 s/vB, a 148 vB P2PKH
> input costs 15 sats, a 68 vB P2WPKH input costs 7 sats, and a 57.5 vB
> P2TR input costs 6 sats. It doesn’t seem obvious to me that saving a few
> dozen sats would greatly foster the users’ urge to consolidate. It feels
> like a lot of overhead for such a small incentive to the users, and
> relying on the miners to give away blockspace below market value feels a
> bit optimistic as well.
>
> Cheers,
> Murch
>
> On 2025-12-11 04:53, defenwycke wrote:
> > Hello list,
> >
> > I’ve been working on a small policy proposal that aims to address one
> > very specific problem: the long-term accumulation of uneconomical dust
> > in the UTXO set.
> >
> > The idea is intentionally narrow. I’m calling it DustSweep, and it
> > defines a strict, non-abusable class of transactions that nodes may
> > relay and miners may include only when the mempool and block space are
> > underutilised. The goal is to give wallets a predictable way to compact
> > dust without introducing new spam vectors or touching consensus.
> >
> > A DustSweep transaction has the following properties:
> >
> > *
> >
> > all inputs are “dust-class” UTXOs
> >
> > *
> >
> > only standard scripts (P2PKH / P2WPKH / P2TR)
> >
> > *
> >
> > exactly one output
> >
> > *
> >
> > no metadata at all (no OP_RETURN, inscriptions, TLVs, etc.)
> >
> > *
> >
> > minimum of 5 inputs (to ensure meaningful UTXO reduction)
> >
> > *
> >
> > size capped
> >
> > *
> >
> > it pays a flat 1 sat per input fee
> >
> > Nodes place these in a small, separate sub-mempool. They’re only
> > accepted when the normal mempool is <50% full, and they’re automatically
> > evicted if normal mempool usage hits 95%. Miners can include them up to
> > a small weight fraction (I suggest ~5%) but only after filling the block
> > with regular fee-paying transactions. The intention is that DustSweep
> > never competes with the fee market and only uses blockspace that would
> > otherwise go unused.
> >
> > This is all policy-level. No consensus changes, no new transaction
> > format, nothing that affects validation. Nodes that don’t implement it
> > simply treat these as low-fee transactions and drop them.
> >
> > The motivation is straightforward: we don’t currently have a safe,
> > structured way to compact dust, and the UTXO set continues to grow from
> > outputs that are effectively unspendable under normal fee conditions.
> > DustSweep tries to offer a predictable, opt-in mechanism for wallets to
> > clean that up without creating any new attack surface.
> >
> > Full draft BIP and supporting documents are here:
> >
> > https://github.com/defenwycke/bip-dust-sweep
> >
> > I’d appreciate feedback on the policy details, thresholds, and whether
> > this fits within what node operators and wallet developers would
> > actually want to use. Happy to adjust parameters if there’s a better
> > balance point.
> >
> > Kind regards,
> >
> > Defenwycke
> >
> > --
> > You received this message because you are subscribed to the Google
> > Groups "Bitcoin Development Mailing List" group.
> > To unsubscribe from this group and stop receiving emails from it, send
> > an email to bitcoindev+...@googlegroups.com
> > <mailto:bitcoindev+...@googlegroups.com>.
> > To view this discussion visit https://groups.google.com/d/msgid/
> > bitcoindev/b47aa182-bca7-44d7-bed1-f3cc2df30ef5n%40googlegroups.com
> > <https://groups.google.com/d/msgid/bitcoindev/b47aa182-bca7-44d7-bed1-
> > f3cc2df30ef5n%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
>
--
You received this message because you are subscribed to the Google Groups "Bitcoin Development Mailing List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bitcoindev+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/841d6882-9853-4d96-84e8-c4742e17e969n%40googlegroups.com.
[-- Attachment #1.2: Type: text/html, Size: 13726 bytes --]
next prev parent reply other threads:[~2025-12-19 1:49 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-12-11 12:53 defenwycke
2025-12-12 18:10 ` [bitcoindev] " Jonathan Voss
2025-12-12 20:17 ` Defenwycke
2025-12-12 22:49 ` [bitcoindev] " Murch
2025-12-13 14:56 ` defenwycke [this message]
2025-12-22 19:06 ` Murch
2025-12-22 19:33 ` Defenwycke
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=841d6882-9853-4d96-84e8-c4742e17e969n@googlegroups.com \
--to=cal.defenwycke@gmail.com \
--cc=bitcoindev@googlegroups.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox