Hi waxwing,
Thanks again for sharing your feedback! The idea of
incrementing the marker byte with a group ID is clever
and I agree that it works technically. This triggered
me to spend some more time to think about what that
might look like which is fun but I am still not
fully convinced this is worth adding.
> Say a bunch of people want maximally weight-minimising
> transaction (for their consolidation), but don't want to
> run a collaborative scheme between each other because
> it's technically hard to do safely (constrained
> hardware/software, nonce state handling, etc.). Each of
> them just want to full agg their 10 inputs, each.
You may have already had this in mind but to make sure
we are on the same page: In this scenario the parties do
not become fully independent of each other. With sighash
default (or all) every signature commits to all outputs,
so all parties have to agree on the complete output list
of the transaction before anyone can sign. Only the
signing session itself is avoided. It is hard for me to
imagine a scenario where that kind of up-front
coordination is fine but the signing session isn't. It
would require some kind of communication and holding of
state as well. Please let me know if you think
differently here. I am not sure what your assumptions
are around up-front coordination capabilities.
I also looked into decoupling the parties further but
it does not really work in my opinion. SINGLE with
ANYONECANPAY only fits parties with one input and one
output each, the consolidation example has no outputs that
the other 9 inputs could safely commit to. When this does
fit, it reveals which input pays which output, and since
nothing commits to the other groups, anyone can split the
merged transaction apart. The explicit sighash bytes
also cost one extra weight unit per input, while
separate transactions can omit the byte by using the
default.
> Not saying it's some super-common-in-practice scenario,
> but it seems suboptimal that you can't do it. So ... I
> guess it depends on whether "the additional validation
> complexity" is really bad compared with the utility of
> that specific use case.
The utility seems limited because of the minor additional
savings they gain from this. They can use half-aggregation
within the shared transaction, but probably better for a
scenario as you outline, each party just broadcasts its
own transaction with its own full-aggregation group. The
latter has about 10 vbytes of transaction overhead and
both alternatives lose all the coordination requirements.
However, we could take the multi-full-agg-group concept
to the next level and half-agg the full-agg signatures
of the different groups for another round of savings (as
well as additional complexity of course).
So, back of the envelope multi-full-agg-group steelman
calculation for your scenario: 10 participants each
consolidating 10 inputs into 1 output.
Each in their own tx:
10 x input skeleton 1640 WU
9 placeholders + 1 final (27 + 67) 94 WU
1 output 172 WU
tx skeleton 42 WU
total 1948 WU (487 vb)
All in one shared tx with half-agg on top:
10 x input skeleton 1640 WU
9 placeholders + 1 R-only final (27 + 35) 62 WU
1 output 172 WU
share of combined s (32 / 10) 3.2 WU
share of tx skeleton (42 / 10) 4.2 WU
total 1881 WU (470 vb)
So this takes it up to 17 vbytes, or 3.4% savings per
participant/group. I didn't fully sketch this out but
I think it should be possible without introducing
additional overhead.
I start to warm up to it because it's clever but I am
not sure if that is a good thing ;)
> More specifically, though, why couldn't a 'group ID' be
> folded into the marker byte.
> Is this just out of consideration for some other reason
> that (hardly surprising) I'm not aware of?
Yepp, this works just like you describe. The reason is
only that I do not see a real use case justifying it and
I even feel like the sighash constraints make reasonable
use-cases hard to imagine.
> A negative (for privacy specifically) would be that you
> tag the different groups in the input list.
Right, the group numbers would reveal which inputs
belong together. But for your consolidation scenario this
seems like a minor issue and it's not that different from
potential finger printing issues resulting from unique
mixing of different aggregation types in a transaction.
I already have some fingerprinting warnings in the BIP to
make people aware of this.
> While that as a proposal may or may not be a good one, I
> feel like: this is worth mentioning at this time because
> (IIUC) it cannot be retrofitted in after.
Yes, undefined marker values cannot become valid through
a soft fork, I thought about something like this early on
trying to find potential upgrade paths. But this doesn't
seem possible and it's good to have this conversation now.
Thank you again for raising this. I have expanded the
rationale section in the BIP with my reasoning so
that the decision is better documented. I am also
still open to think about other potentially interesting
use-cases for the multi-group-full-agg feature or to
reconsider if you or other reviewers think we should rather
be maximimally flexible in this regard even without a clear
use-case. But unless there is interest being signalled
I will err on the side of simplicity.
Best,
Fabian
On Wednesday, July 22nd, 2026 at 2:21 AM, waxwing/ AdamISZ <
ekag...@gmail.com> wrote:
Hi Fabian,
Thanks again for the draft.
"A more flexible alternative with multiple aggregation groups per scheme,
similar to the bucket concept from early aggregation discussions, was
rejected because no concrete use case justified the additional
validation complexity."
Say a bunch of people want maximally weight-minimising transaction (for their consolidation), but don't want to run a collaborative scheme between each other because it's technically hard to do safely (constrained hardware/software, nonce state handling, etc.). Each of them just want to full agg their 10 inputs, each.
Not saying it's some super-common-in-practice scenario, but it seems suboptimal that you can't do it. So ... I guess it depends on whether "the additional validation complexity" is really bad compared with the utility of that specific use case.
More specifically, though, why couldn't a 'group ID' be folded into the marker byte. In your draft you say "The chosen values fall into the range 0xbb to 0xfe, which is unassigned in legacy script and corresponds to OP_SUCCESS opcodes in tapscript, so they do not collide with opcodes commonly seen in scripts." - but there are a lot of values between 0xbd and 0xfe. I guess this is not a case for bitflags, but just different markers where you add a groupID to 0xbd (and with the understanding - this clearly is only relevant for fullagg, not halfagg). Is this just out of consideration for some other reason that (hardly surprising) I'm not aware of?
(I don't believe this suggestion changes anything functionally as described in the 'Common Signature Message' section; I mean it literally changes the byte, but, functionally).
A negative (for privacy specifically) would be that you tag the different groups in the input list.
While that as a proposal may or may not be a good one, I feel like: this is worth mentioning at this time because (IIUC) it cannot be retrofitted in after.
Cheers,
AdamISZ/waxwing
On Saturday, July 18, 2026 at 11:44:33 AM UTC-3 Fabian wrote:
Hi list,
I would like to share a BIP draft for transaction-wide cross-input
signature aggregation (CISA). It introduces a new witness version,
which enables Taproot-style key path spending where inputs can
aggregate their signatures.
Each input chooses between half-aggregation, full-aggregation, or an
explicit opt-out via a marker byte in its witness.
The aggregation schemes themselves are specified in the previously
shared BIP 458 (half-aggregation) and BIP 459 (full-aggregation)
drafts. Script path spending follows BIP 341/342 unchanged.
The BIP draft text can be found here:
For inline comments, I have opened a mock pull request on my BIPs
repo fork:
Feedback of any kind is much appreciated.
Best,
Fabian
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.