RFC: Do not RBF external transactions by default #36605

issue Eunovo opened this issue on October 22, 2026
  1. Eunovo commented at 10:55 AM on October 22, 2026: contributor

    External transactions refer to transactions received via the p2p network as opposed to transactions created and saved via the send RPCs.

    The existence of silent payments transactions creates a new challenge for RBF. The addition of a new input to the input set of the original transaction, requires regeneration of the original transaction's silent payments outputs for the replacement transaction. This process requires the original SP payment instructions; #35302 persists these payment instructions for this purpose. However, if the transaction being replaced is an "external" transaction, the payment instructions are not available to the wallet.

    There is no way to identify silent payments transactions, so a possile solution is to flag transactions that are external and require that the user explicitly force the replacement of such transactions. A new parameter should be added to the bumpfee RPC that must be set to true by the user to allow the replacement of an external transaction. In addition to this, a new parameter should be added to the bumpfee RPC to allow SP payment instructions to be provided.

  2. Eunovo commented at 9:20 AM on October 29, 2026: contributor
  3. bartoli commented at 4:35 PM on October 29, 2026: none

    Maybe stick with a single option 'sp_info' or something, which would be the list of sp addresses of each output, or an empty list for regular transactions? That way, caller is aware that he has to think about sp payments, and it feels less aggressive than forcing 99% of the use cases to set force=true everytime?

  4. Eunovo commented at 9:42 AM on October 30, 2026: contributor

    Maybe stick with a single option 'sp_info' or something, which would be the list of sp addresses of each output, or an empty list for regular transactions?

    I'm not sure this is enough to make users actively think about SP payments unless you are suggesting that the sp_info parameter be required?

    and it feels less aggressive than forcing 99% of the use cases to set force=true everytime?

    Where does the 99% figure come from? I’d expect most users not to have the same keys loaded in multiple wallets. The main exception is setups that pair an offline signer with an online watch-only wallet.

  5. bartoli commented at 10:18 AM on October 30, 2026: none

    3. I'm not sure this is enough to make users actively think about SP payments unless you are suggesting that the sp_info parameter be required?

    Yes, i was thinking about making this one mandatory, which seems clearer than having a 'force' argument that has to be here for all non-silent payment txs without having a name giving context what it is for

    3. Where does the 99% figure come from? I’d expect most users not to have the same keys loaded in multiple wallets. The main exception is setups that pair an offline signer with an online watch-only wallet.

    That was my rough idea that ~1% of bumpfee calls would be for silent payments once the feature is merged. Actually, I think it would be the main pain point of integrating this: forcing almost everyone to add an argument they don't use, just because the node can't know if an external transaction contains silent payments?

    Also, I was actually expecting the bumpfee could only be used for silent payments if the tx is paid by a wallet known by the node receiving the rpc request? But maybe the SP instructions can transmit the new S value or something like that without making the sender's private keys computable by another node?

    I might not have fully understood the request yet, but I would be interested in working on it

  6. Eunovo commented at 11:53 AM on October 30, 2026: contributor

    That was my rough idea that ~1% of bumpfee calls would be for silent payments once the feature is merged. Actually, I think it would be the main pain point of integrating this: forcing almost everyone to add an argument they don't use, just because the node can't know if an external transaction contains silent payments?

    Most users shouldn't have the same descriptors loaded in multiple wallets; the only valid setups I can think of are offline signer/online watch-only wallet pairs. I expect most wallets to have few external transactions, but I could be wrong.

    Also, I was actually expecting the bumpfee could only be used for silent payments if the tx is paid by a wallet known by the node receiving the rpc request? But maybe the SP instructions can transmit the new S value or something like that without making the sender's private keys computable by another node?

    I'm a bit confused here. SP txs are like regular txs and can be fee-bumped in the usual way. The idea here is to prevent users from accidentally burning UTXOs.

    Yes, i was thinking about making this one mandatory, which seems clearer than having a 'force' argument that has to be here for all non-silent payment txs without having a name giving context what it is for

    I think it might be weirder to have a mandatory sp_info argument. It is less weird if it is only mandatory for external transactions as opposed to all transactions. Maybe this is what you meant? I can see how just having sp_info and requiring it for external transactions can be better than requiring force=True with the optional sp_info.

  7. bartoli commented at 12:30 PM on October 30, 2026: none

    5. I'm a bit confused here. SP txs are like regular txs and can be fee-bumped in the usual way. The idea here is to prevent users from accidentally burning UTXOs.

    I just mean, in the case SP outputs have to be recomputed by the node receiving the feebump RPC (because as said in the issue description, inputs would change), either the node has to have a wallet knowing the private keys of the wallet owning the utxos being spent, or SP instructions have to provide a way to recompute the new SP outputs, but without giving enough info to be able to recompute the sender's private key.

    Maybe you already accounted for this and there is no problem.

  8. Eunovo commented at 9:54 AM on November 2, 2026: contributor

    6. either the node has to have a wallet knowing the private keys of the wallet owning the utxos being spent

    Yes, the wallet being considered in this issue has the keys to the received transaction; hence, it can fee-bump it.

  9. achow101 commented at 2:41 AM on November 3, 2026: member

    Since sendrawtranasction does not operate on wallets, this would mean that anything that is broadcast through sendrawtransaction is treated as external.

  10. Eunovo commented at 9:45 AM on November 3, 2026: contributor

    8. Since sendrawtranasction does not operate on wallets, this would mean that anything that is broadcast through sendrawtransaction is treated as external.

    It should be limited to transactions that the wallet can fee-bump; transactions that contain inputs the wallet has keys for.

  11. polespinasa commented at 10:49 AM on November 3, 2026: member

    It is tricky problem, I wonder what are the other wallets doing for this issue. How are Sparrow and others addressing this issue?

    If we go with the main idea proposed by this issue, probably the mandatory signal of that extra parameter could be limited to transactions with P2TR outputs. We should probably add also some way to export-import SP information between wallets, some export/import address book between wallets which I guess would require a BIP in order to standardize it between wallets..

  12. bartoli commented at 3:45 PM on November 4, 2026: none

    The only thing I could find is that Cake Wallet disables speeding up silent payments. But that means they are talking about 'internal' ones they know the destination sp address of.

    I'm wondering if a 'less tricky' thing would be to start by reviving the issues like #20935, which already had some PR started :https://github.com/bitcoin/bitcoin/pull/25776 So when calling bumpfee, user could still at least fee bump silent payments until the current change output allows, by just adding an option 'do not add inputs'

    Of course in this case (blindly bumping without recalculating SP outputs), we would also have to make sure other things that could break the SP outputs don't occur, like guaranteeing outputs keep the same order


github-metadata-mirror

This is a metadata mirror of the GitHub repository bitcoin/bitcoin. This site is not affiliated with GitHub. Content is generated from a GitHub metadata backup.
generated: 2026-11-16 01:51 UTC

This site is hosted by @0xB10C
More mirrored repositories can be found on mirror.b10c.me