On Wednesday, June 17th, 2026 at 11:20 PM, Anthony Towns aj@erisian.com.au wrote:

On Tue, Jun 16, 2026 at 08:09:08PM +0000, Pieter Wuille wrote:

I want to first correct a potential misunderstanding here, because
I realize the terms "Later" and "Never" aren't very descriptive. They
are specifically about an expected and relied-upon expectation that an
EC-disabling fork will happen that at least applies to the output type
itself, in time. "Later" is the expectation that such a disabling will
happen after the output type is introduced, but still in time (so, before
Q-day). Outputs without a strong expectation that their EC paths/opcodes
will be disabled, or not in time, I classify under "Never".

I believe here you're instead arguing for P2MR ("Merkle-Never")
over all "Later" options. That was my previous point: I think (solely)
having "Never" output types like P2MR is just utterly insufficient for
any worthwhile migration. It's so incompatible with today's workflows
that it either won't be adopted, or (possibly inadvertently) adopted
in an insecure fashion. Yes, it gives people the option to safeguard
their own coins, but to me that's disaster recovery territory - I think
we ought to prioritize maximizing the chances for saving the currency
as a whole in case Q-day comes, not a small subset of individual users'
coins. P2MR (alone) doesn't really achieve much in that regard in my view,
thus we at least need something of the "Later" class in addition.

I'm not sure I follow/agree with the logic here. I think the general idea
is:

1) we create some new address types that allow post-quantum spending, but
also allow efficient quantum-vulnerable spending that can be used prior
to Q-day

2) many people migrate to these new address types

3) Q-day arrives

4) all spending goes via the post-quantum paths, and everyone's funds are
safe

There are three main variations to this, I think:

Later-dominant: towards the end of (2) but prior to (3), the
quantum-vulnerable spend paths are disabled in a predictable, planned
manner, preventing coin theft, but not disrupting active usage
significantly (or not disrupting it any more than the proximity of
Q-day already is).

Self-reliance: Q-day goes from maybe to definitely faster than consensus
changes can be coordinated, and the only reason people's funds remain
safe is that they can (and do) keep the quantum-vulnerable spend
paths secret.

Disaster-recovery: neither the "predictable/planned consensus change" of
Later nor the "everyone takes responsiblity for their own funds"
works, and the only way to avoid a large percentage of bitcoin
becoming a reward for quantum research is to replace EC spend paths
with a ZK proof of knowledge of a BIP32 seed or similar, requiring
a hard fork. Such a hard fork would be certain to be controversial
("why at this height: I received a payment five blocks after //
my funds were stolen by a quantum attacker five blocks earlier //
why are only BIP32 funds recoverable not scheme X"), but if no other
approach works, might still be the ultimately outcome.

So the point here was just: if you already accept the need for a "Later"
output type (= one with builtin-in EC disabling expected from the start),
P2TRv2 is preferable over P2MR+ED, because:

As far as I can see, only having P2TRv2 addresses would rule out the
"self-reliance" path here.

Picking when Q-day will occur, either individually for determining
your own security posture, or collectively for organising a consensus
change seems very difficulty: it involves evaluating both complex state
of the art physics research, but also estimating the covert abilities
of national governments and the incentives to attack bitcoin prior to
revealing their capabilities. To me, that doesn't bode well for a smooth
and fast deployment of a consensus change in advance of problems occuring.

It's possible that general carelessness on behalf of users would also
rule out the effectiveness of a self-reliance approach: if only the most
conscientious 1% of users make use of P2MR securely, that might secure 10%
of funds, but having 90% of the bitcoin supply crash probably wouldn't be
very valuable either. But even then, that may make the "disaster-recovery"
approach less problematic, by ensuring the 1%/10% who were conscientious
can avoid being part of the "my funds were stolen by a quantum attacker"
contingent, eg.

Theorycrafting for a second here. It's reasonable to conjecture fee
rates will be much higher post-Q-day, and thus P2MR's 32 byte advantage
over P2TRv2 will yield significant savings for end-users in absolute
terms. If fee rates inflate 10x higher after Q-day, then 8 vbytes becomes
significant (100 sats per spend or more).

I don't think that is the right way to look at. 8vb/input is about
an additional 14% today (vs a taproot key-path spend), but with the
post-quantum signatures we have available now that's likely to reduce
to ~7% (SHRINCS) or ~1% (winternitz). So, post-Q-day, by dropping 32B,
you're only getting an expected savings in fees / increase in block
capacity on that order of magnitude, ie: 1%-7%. That's nice to have,
for sure, but doesn't compare to introducing a new address type that
puts the PQ sigs in an extension block, or one that uses ZK proofs to
do cross-input or cross-transaction signature aggregation, eg. So a 32B
witness overhead for an unused/unusable key-path post-Q-day doesn't seem
very relevant to me.

FWIW, I don't think the P2TRv2 EC-disabling fork needs to be timed
perfectly. The expectation should be just that it happens before Q-day,
and when it looks inevitable or the fear about that is large enough.

FWIW, I would define "timed perfectly" precisely as "EC disabling
fork happens before Q-day". Maybe allowing some additional months of
EC-efficiency would be a win while not taking out too much migration time,
but for me "perfection" here means "no one who upgraded lost money via
quantum-related vulnerabilities".

One reason I'm doubtful is that I think for some the "it looks inevitable"
threshold has already been crossed, eg:

So let me attempt to partially fill the silence, similarly to what
Scott Aaronson did in his April 29 post. Given everything I know,
including scary non-public information, I now put the odds of qday by
2032 at 50%. 10% by 2030.

IMO a good target date for migration is 2029, roughly 3.5 years
out. 2029 happens to be the date selected by Google, Cloudflare, and
the Ethereum Foundation.

https://x.com/drakefjustin/status/2061793725299224676

So, here it is: if quantum computers start breaking cryptography a
few years from now, don’t you dare come to this blog and tell me that
I failed to warn you. This post is your warning. Please start switching
to quantum-resistant encryption, and urge your company or organization
or blockchain or standards body to do the same.

https://scottaaronson.blog/?p=9718

Personally, that leaves me at a minimum very skeptical of the utility
of introducing either P2MR or P2TRv2 (etc) approaches that don't also
introduce a quantum-safe spending path on day one.

First a meta-point here: the reason I like separating the discussion into different dimensions than just "P2TRv2 vs P2MR" is because there are more options than those two, and not every argument applies to the same separation into two classes. Let me list them explicitly here, roughly in decreasing order of how strongly I feel about them:
- We need at least a "Later" option for meaningful migration, i.e. P2TRv2 or P2MR+ED.
- Within the "Later" option, P2TRv2 is preferable.
- A "Later" option benefits from being the only PQC migration strategy, making it a Schelling point.

Correct me if I'm mistaken, but I think P2TRv2 is preferable only in the
"Later-dominant" scenario, and only to the degree that it's slightly
cheaper prior to Q-day. If it were the only available option, that would
increase the risk of loss involved with both the other approaches. (I
don't think P2TRv2 is meaningfully more private than P2MR in the way
taproot v1 is, as presumably you'd only adopt that address format if
you wanted to have a post-quantum script path)

You'd have to convince everyone to deploy and then adopt P2TRv2 today under the public knowledge that it is insecure and their coins are more likely to be stolen. That's a hard sell.

How does one pitch P2TRv2 to users? "It will be quantum secure one day we promise because everyone is incentivized to fix it later as Bitcoin will die if we don't."

How do you pitch P2MR? "It's quantum secure if you use it correctly."
To me, the pitch is "Bitcoin can only remain valuable if we mostly/all migrate." for both. P2TRv2 is just much easier to adopt, because P2MR (or any "Never" output type) fundamentally requires many users to change their workflows entirely.

Let's call NoEC-day the earlier of activation of a soft-fork disabling
EC-spends on P2MR/P2TRv2 or Q-day. NoEC-day to some extent is presumably
equal to "the day the bitcoin ecosystem has finally agreed that CRQCs
are inevitable".

My (cynical?) view is the only people who will adopt either P2TRv2 or
P2MR prior to NoEC-day being schedule will be people who are willign
to use those features in a quantum-safe way -- that is, keeping their
EC pubkeys secret, and only revealing those EC pubkeys to spend them
immediately, prior to NoEC-day. In that view, the EC-spend-paths of such
coins prior to NoEC-day are only an opportunistic "make spends cheaper"
shortcut, they don't allow the funds to be used in lightning channels
or to let your wallet be audited via sharing an xpub or anything similar.

Perhaps I'm wrong: it's my opinion, not a technical fact; it's possible
that lightning software could get an upgrade to generate post-quantum
signatures for channel commitments or HTLC claims, I just think it's
pretty unlikely that that will happen at a meaningful scale when everyone
has much more immediate and less theoretical things to work on prior to
NoEC-day, especially when the efficiency/performance of such changes is
likely to be very low.

This focus on allowing individual users the ability to safeguard their
coins just feels like a red herring: [...]

While I appreciate your point, I also feel that "allowing individual
users the ability to safeguard their coins" is pretty close to the entire
point of Bitcoin. :)

In either case, I consider anything that requires hardcoding
specific wallet designs (BIP32 or otherwise) into Bitcoin's consensus
rules (and only allowing those coins to survive) to be squarely in
disaster-recovery land. It feels like embracing arbitrariness, and
giving up on the permissionlessness that makes Bitcoin interesting -
if the community shows it can get consensus on effectively burning
coins except those that match a whitelist, it feels hard to maintain
censorship-freeness as a value, even if the whitelist includes most of
the (active) coins. That is of course not to say such techniques aren't
interesting to work on, but to me, their place is in disaster recovery
scenarios to save what's left, after Q-day, if migration attempts were
unsuccessful. In such a setting, I think we're already in effectively an
altcoin-with-UTXO-bootstrapped-from-Bitcoin territory, and a (possibly
growing) set of ways to recover burned coins can be hardforked in.

Just for the record, I think the above is an accurate description of the
"disaster-recovery" scenario above: the "quantum-safe" hard-fork variant
of bitcoin would be fairly described as a utxo-bootstrapped altcoin,
would compete in the market with the "quantum-unsafe" bitcoin that
existing nodes would continue to follow, and possibly there would be
many slightly varying such altcoins competing with each other, eg on
exactly what height the utxo snapshot was taken or what coins become
spendable. It would not be a fun time for holders of affected utxos.

My impression is that your approach is to have an answer for many
possible futures, including ones where Q-day arrives within just a few
years.

"The key of strategy... is not to choose a path to victory, but to choose
so that all paths lead to a victory."

-- https://tvtropes.org/pmwiki/pmwiki.php/Main/XanatosGambit

But optimizing for disaster-recovery also means reducing the
chances of preserving Bitcoin as we know it in the scenarios where a
successful migration is still possible. And if Q-day does arrive that
soon, I don't see what we can do today that would preserve Bitcoin in
a form we care about anyway. By accepting that, we can focus on the
futures where our choices today can still materially improve the outcome.

Preserving bitcoin as a personally-possessible inflation resistant
store of value seems both possible and worth caring about, even if other
constraints means that many people can't afford to personally hold it
(and have to go through ETFs/exchanges/banks) and that it can't be used
for day-to-day transactions. Would be very disappointing if true, and
even given Q-day in a few years I expect we could do better than just
that, but it doesn't feel like a throw-in-the-towel event to me.

Essentially yes though, I expect the majority of holders will probably
migrate to PQ addresses via rescue protocols, either on Bitcoin or a fork
thereof. Even if we can't come to consensus and deploy a new output type,
we'll still be able to rescue most coins. It's just that we'd have nowhere
to rescue them to, so we ought to make PQ wallets available soon, so we're
not in a rush to build them later when we need them. If a PQ wallet can
use cheap EC signatures while they're still trustworthy, all the better

FWIW, that's my guess on how things would play out if the near-term Q-day
timelines I've seen (ie, 2029 to 2035) match reality. I hope that's
pessimistic (either because the Q-day timelines are bad estimates, or
because migration happens in a more orderly fashion), but I guess we'll
see. I don't rate my ability to evaluate Q-day predictions very highly.

- A (not-quite-CR)-QC breaks 128-bit ECDLP, say.

I'm not in a position to judge, but the google paper claims:

"Indeed, if a leading quantum architecture encounters and overcomes
all its scaling challenges before producing a device able to
solve (for example) 32-bit ECDLP, then there may be little time
between the breaking of 32-bit ECDLP and the breaking of 256-bit
ECDLP. Furthermore, the community should not expect to see published
demonstrations of the most advanced quantum error-correction
architectures and quantum algorithms deployed to cryptanalytic
problems. Thus, a successful public demonstration of Shor’s
algorithm on a 32-bit elliptic curve should not be seen as a wake-up
call to adopt PQC as much as a potential signal that PQC adoption
has already failed."

and places the required tiffoli gates and logical qubits for a 32-bit
break at about (300k, 200) vs (10M, 600) for 128-bit or (80M, 1100)
for 256-bit.

Of course, if you believe it's the only possible future, I understand
you'd come to a different conclusion. But is it really? Do you think
this isn't a plausible future:

- A P2TRv2 type (let's leave aside whether P2MR or P2QR gets added too) gets introduced in the next few years, with hash-based PQC opcodes.
- Over the course of the next decade or so, it gets adopted by practically all active users.

I think it might be better to look at that scenario in a more fine-grained
way? I think your "Late-ish" scenario is:

1) P2TRv2 (or whatever) is introduced
2) Some PQ opcodes get enabled, allowing expensive but PQ-safe spend-paths
via those outputs
3) P2PK, P2PKH, P2WPKH, P2WSH, P2TR all become obsolete in favour of P2TRv2,
but EC spend paths continue to be what's used in practice.
4) Some better PQ solutions become available, allowing cheap PQ-safe spend-paths
5) Everyone switches from EC paths to the new PQ paths.
6) NoEC-day happens, but no one is impacted.

I think your "Soon-ish" scenario differs as of step (4):

4) NoEC-day happens. Massive disruption because the "what's used in practice"
path breaks, but everything is recoverable.
5) Post-quantum approaches get even higher priority

I'm skeptical of step 3 here; and would expect to see something more like:

3') Only a small proportion of users (ie, the most conscientious/fearful)
switch to P2TRv2 with most preferring to stick with what works

That has no real impact on the Late-ish scenario, but changes the Soon-ish
scenario to either:

4'a) NoEC-day happens substantially before Q-day; people hurry to migrate
to P2TRv2, but it mostly works.

or

4'b) NoEC-day happens essentially at the same time as Q-day; coins get
stolen and we hit the disaster-recovery scenario.

Perhaps the difference between (3') and (3) playing out in reality
is just having nearly everyone agree that the upgrade is essential,
and rather than leaving it to self-interest/market-dynamics, having a
consistent push that every single wallet/service that doesn't deprecate
every current address type is a danger to the entire ecosystem, even
absent widespread agreement on when/if Q-day will happen. Arguably that
would be easier to agree on if the incremental cost of EC spend paths
in P2TRv2 prior to NoEC-day/Q-day versus current spend paths is near to
zero, rather than up to ~14% per input.

Cheers,
aj

--
You received this message because you are subscribed to a topic in the Google Groups "Bitcoin Development Mailing List" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bitcoindev/p8AVEmAtWdA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to bitcoindev+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/bitcoindev/ajN9be00Br-O3-9R%40erisian.com.au.