A BIP proposal for transactions that are 'cancellable'

HASH160 <revokehash> EQUAL
IF
     <Bob's pubkey>
ELSE
     "24h" CHECKSEQUENCEVERIFY DROP
     <Alice's pubkey>
ENDIF
CHECKSIG
HASH160 <revokehash> EQUAL
IF
     <Bob's pubkey>
ELSE
     "-24h" CHECKSEQUENCEVERIFY DROP
     <Alice's pubkey>
ENDIF
CHECKSIG
>Hello everyone,
>
>We would like to propose a new BIP to extend OP_CSV (and/or OP_CLTV) in
>order for these to allow and interpret negative values. This way,
>taking the example shown in BIP 112:
>
>HASH160 <revokehash> EQUAL
>IF
>     <Bob's pubkey>
>ELSE
>     "24h" CHECKSEQUENCEVERIFY DROP
>     <Alice's pubkey>
>ENDIF
>CHECKSIG
>
>that gives ownership only to Bob for the first 24 hours and then to
>whichever spends first, we basically propose using the negative bit
>value:
>
>HASH160 <revokehash> EQUAL
>IF
>     <Bob's pubkey>
>ELSE
>     "-24h" CHECKSEQUENCEVERIFY DROP
>     <Alice's pubkey>
>ENDIF
>CHECKSIG
>
>meaning that both would have ownership for the first 24 hours, but
>after that only Bob would own such coins. Its implementation should
>not be too tedious, and in fact it simply implies considering negative
>values that are at the moment discarded as for the specification of
>BIP-112, leaving the sign bit unused.
>
>This, we argue, an increase the fairness of the users, and can at times
>be more cost-effective for users to do rather than trying a
>Replace-By-Fee
>transaction, should they want to modify such payment.
>
>We would like to have a discussion about this before proposing the
>BIP, for which we are preparing the text.
>
>You can find our paper discussing it here:
>https://hal-cea.archives-ouvertes.fr/cea-01867357 (find attached as
>well)
>
>Best,
>
>--
>Alejandro Ranchal Pedrosa, Önder Gürcan and Sara Tucci-Piergiovanni
> Hello everyone,
>
> We would like to propose a new BIP to extend OP_CSV (and/or OP_CLTV) in
> order for these to allow and interpret negative values. This way,
> taking the example shown in BIP 112:
>
> HASH160 <revokehash> EQUAL
> IF
>      <Bob's pubkey>
> ELSE
>      "24h" CHECKSEQUENCEVERIFY DROP
>      <Alice's pubkey>
> ENDIF
> CHECKSIG
>
> that gives ownership only to Bob for the first 24 hours and then to
> whichever spends first, we basically propose using the negative bit value:
>
> HASH160 <revokehash> EQUAL
> IF
>      <Bob's pubkey>
> ELSE
>      "-24h" CHECKSEQUENCEVERIFY DROP
>      <Alice's pubkey>
> ENDIF
> CHECKSIG
>
> meaning that both would have ownership for the first 24 hours, but
> after that only Bob would own such coins. Its implementation should
> not be too tedious, and in fact it simply implies considering negative
> values that are at the moment discarded as for the specification of
> BIP-112, leaving the sign bit unused.
>
> This, we argue, an increase the fairness of the users, and can at times
> be more cost-effective for users to do rather than trying a Replace-By-Fee
> transaction, should they want to modify such payment.
>
> We would like to have a discussion about this before proposing the
> BIP, for which we are preparing the text.
>
> You can find our paper discussing it here:
> https://hal-cea.archives-ouvertes.fr/cea-01867357 (find attached as well)
>
> Best,
>
> --
> Alejandro Ranchal Pedrosa, Önder Gürcan and Sara Tucci-Piergiovanni
>
> _______________________________________________
> bitcoin-dev mailing list
> bitcoin-dev@lists•linuxfoundation•org
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>
> Dear Matt,
>
> Notice that what you suggest has some substantial differences. With your
> suggestion of a multisig option with a 24h timelock, once you give Alice
> the chance to spend that UTXO without a negative timelock (as we argue),
> by means of, say, a transaction that she can use, you cannot enforce
> that this is not used by Alice after the 24hs. Perhaps it is possible,
> tweaking the Lightning Channel design of Breach Remedy txs, to penalize
> Alice if she does this, but this requires Bob to check the Blockchain in
> case he needs to publish a proof-of-fraud, think of adding extra funds
> to the transaction to account for penalization, etc.
>
> Feel free to correct me if I got it wrong in your email.
>
> Best,
> Alejandro.
>
>
> On Thu, Sep 6, 2018 at 3:32 PM Matt Corallo <lf-lists@mattcorallo•com
> <mailto:lf-lists@mattcorallo.com>> wrote:
>
>     I think a simple approach to what you want to accomplish is to
>     simply have a multisig option with a locktime pre-signed transaction
>     which is broadcastable at the 24h mark and has different
>     spendability. This avoids introducing reorg-induced invalidity.
>
>     On September 6, 2018 9:19:24 AM UTC, Alejandro Ranchal Pedrosa via
>     bitcoin-dev <bitcoin-dev@lists•linuxfoundation•org
>     <mailto:bitcoin-dev@lists.linuxfoundation.org>> wrote:
>
>         Hello everyone,
>
>         We would like to propose a new BIP to extend OP_CSV (and/or OP_CLTV) in
>         order for these to allow and interpret negative values. This way,
>         taking the example shown in BIP 112:
>
>         HASH160 <revokehash> EQUAL
>         IF
>              <Bob's pubkey>
>         ELSE
>              "24h" CHECKSEQUENCEVERIFY DROP
>              <Alice's pubkey>
>         ENDIF
>         CHECKSIG
>
>         that gives ownership only to Bob for the first 24 hours and then to
>         whichever spends first, we basically propose using the negative bit value:
>
>         HASH160 <revokehash> EQUAL
>         IF
>              <Bob's pubkey>
>         ELSE
>              "-24h" CHECKSEQUENCEVERIFY DROP
>              <Alice's pubkey>
>         ENDIF
>         CHECKSIG
>
>         meaning that both would have ownership for the first 24 hours, but
>         after that only Bob would own such coins. Its implementation should
>         not be too tedious, and in fact it simply implies considering negative
>         values that are at the moment discarded as for the specification of
>         BIP-112, leaving the sign bit unused.
>
>         This, we argue, an increase the fairness of the users, and can at times
>         be more cost-effective for users to do rather than trying a Replace-By-Fee
>         transaction, should they want to modify such payment.
>
>         We would like to have a discussion about this before proposing the
>         BIP, for which we are preparing the text.
>
>         You can find our paper discussing it here:
>         https://hal-cea.archives-ouvertes.fr/cea-01867357 (find attached as well)
>
>         Best,
>
> Functionality such as this does not currently exist not because no one
> thought of it before, but because it has been proposed many times
> before and determined to be harmful.  The existing design of CLTV/CSV
> were carefully constructed to make it impossible for a transaction to
> go from valid to invalid based on the time. The most naive
> construction-- e.g. push the current time/height on the stack-- would
> have that property and was specifically avoided.
>
> When a spend goes from valid to invalid it means that a reorganization
> will destroy coins even completely absent any dishonest actions of the
> coins prior owner in the coins recent casual history. Effectively a
> coin with any kind of non-monotone validity event in its recent
> history functions like a recently generated coin: a coin that reorgs
> destroy. Bitcoin addresses the issue for recently generated coins by
> not permitting their use for 100 blocks.  I've yet to see an argument
> for a use case for non-monotone validity that still sounds useful once
> the negative effects are addressed (e.g. by subjecting coins that have
> gone through them to a maturity limitation).
> _______________________________________________
> bitcoin-dev mailing list
> bitcoin-dev@lists•linuxfoundation•org
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
> I made a similar proposal about 7 months ago, and documented some of the
> discussion points here:
>
> https://github.com/reardencode/bips/blob/reverselocktime/bip-0zzz.
> mediawiki
>
> On 2018-09-06 (Thu) at 15:16:34 +0000, Gregory Maxwell via bitcoin-dev
> wrote:
> > Functionality such as this does not currently exist not because no one
> > thought of it before, but because it has been proposed many times
> > before and determined to be harmful.  The existing design of CLTV/CSV
> > were carefully constructed to make it impossible for a transaction to
> > go from valid to invalid based on the time. The most naive
> > construction-- e.g. push the current time/height on the stack-- would
> > have that property and was specifically avoided.
> >
> > When a spend goes from valid to invalid it means that a reorganization
> > will destroy coins even completely absent any dishonest actions of the
> > coins prior owner in the coins recent casual history. Effectively a
> > coin with any kind of non-monotone validity event in its recent
> > history functions like a recently generated coin: a coin that reorgs
> > destroy. Bitcoin addresses the issue for recently generated coins by
> > not permitting their use for 100 blocks.  I've yet to see an argument
> > for a use case for non-monotone validity that still sounds useful once
> > the negative effects are addressed (e.g. by subjecting coins that have
> > gone through them to a maturity limitation).
> > _______________________________________________
> > bitcoin-dev mailing list
> > bitcoin-dev@lists•linuxfoundation•org
> > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
> _______________________________________________
> bitcoin-dev mailing list
> bitcoin-dev@lists•linuxfoundation•org
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>
> I think you misunderstood my proposal. What you'd do is the transaction
> is spendable by either Bob OR (Bob AND Alice) and before
> broadcast/during construction/whatever sign a new transaction that
> spends it and is only spendable by Alice, but is timelocked for 24
> hours. At the 24h mark, Alice broadcasts the transaction and once it is
> confirmed only Alice can claim the money.
>
> On 09/06/18 10:59, Alejandro Ranchal Pedrosa wrote:
>> Dear Matt,
>>
>> Notice that what you suggest has some substantial differences. With your
>> suggestion of a multisig option with a 24h timelock, once you give Alice
>> the chance to spend that UTXO without a negative timelock (as we argue),
>> by means of, say, a transaction that she can use, you cannot enforce
>> that this is not used by Alice after the 24hs. Perhaps it is possible,
>> tweaking the Lightning Channel design of Breach Remedy txs, to penalize
>> Alice if she does this, but this requires Bob to check the Blockchain in
>> case he needs to publish a proof-of-fraud, think of adding extra funds
>> to the transaction to account for penalization, etc.
>>
>> Feel free to correct me if I got it wrong in your email.
>>
>> Best,
>> Alejandro.
>>
>>
>> On Thu, Sep 6, 2018 at 3:32 PM Matt Corallo <lf-lists@mattcorallo•com
>> <mailto:lf-lists@mattcorallo.com>> wrote:
>>
>>      I think a simple approach to what you want to accomplish is to
>>      simply have a multisig option with a locktime pre-signed transaction
>>      which is broadcastable at the 24h mark and has different
>>      spendability. This avoids introducing reorg-induced invalidity.
>>
>>      On September 6, 2018 9:19:24 AM UTC, Alejandro Ranchal Pedrosa via
>>      bitcoin-dev <bitcoin-dev@lists•linuxfoundation•org
>>      <mailto:bitcoin-dev@lists.linuxfoundation.org>> wrote:
>>
>>          Hello everyone,
>>
>>          We would like to propose a new BIP to extend OP_CSV (and/or OP_CLTV) in
>>          order for these to allow and interpret negative values. This way,
>>          taking the example shown in BIP 112:
>>
>>          HASH160 <revokehash> EQUAL
>>          IF
>>               <Bob's pubkey>
>>          ELSE
>>               "24h" CHECKSEQUENCEVERIFY DROP
>>               <Alice's pubkey>
>>          ENDIF
>>          CHECKSIG
>>
>>          that gives ownership only to Bob for the first 24 hours and then to
>>          whichever spends first, we basically propose using the negative bit value:
>>
>>          HASH160 <revokehash> EQUAL
>>          IF
>>               <Bob's pubkey>
>>          ELSE
>>               "-24h" CHECKSEQUENCEVERIFY DROP
>>               <Alice's pubkey>
>>          ENDIF
>>          CHECKSIG
>>
>>          meaning that both would have ownership for the first 24 hours, but
>>          after that only Bob would own such coins. Its implementation should
>>          not be too tedious, and in fact it simply implies considering negative
>>          values that are at the moment discarded as for the specification of
>>          BIP-112, leaving the sign bit unused.
>>
>>          This, we argue, an increase the fairness of the users, and can at times
>>          be more cost-effective for users to do rather than trying a Replace-By-Fee
>>          transaction, should they want to modify such payment.
>>
>>          We would like to have a discussion about this before proposing the
>>          BIP, for which we are preparing the text.
>>
>>          You can find our paper discussing it here:
>>          https://hal-cea.archives-ouvertes.fr/cea-01867357 (find attached as well)
>>
>>          Best,
>>
> ade a similar proposal about 7 months ago, and documented some of the
> discussion points here:
> Hi all,
>
> Thank you for the link, and also to Gregory for the remarks. I did not
> know about this previous proposal. I think the last paragraph of future
> work is interesting:
>
> "It may be interesting to add enhance OP_CHECKSEQUENCEVERIFY
> <https://github.com/bitcoin/bips/blob/master/bip-0112.mediawiki> to
> allow outputs that are spendable by Alice until time foo, always
> spendable by Bob, and spendable by Joe only after time bar, or other
> such cases"
>
> Perhaps it would allow this functionality, while keeping the validity of
> coins, if the new OP_zzz took an additional argument than suggested,
> such that the first one is the timelimit for Alice to keep the coin (say
> in the first 24 hours), and after those 24 hours the ownership goes to
> the third argument, say Bob.
>
> That is, it is not possible to use only specifying the owner in the
> first 24 hours. Would this be considered harmful?
>
> Best,
>
> Alejandro.
>
> On 9/6/18 10:32 PM, Brandon Smith wrote:
> > ade a similar proposal about 7 months ago, and documented some of the
> > discussion points here:
Hello Gregory, all
  Thank you so much for your feedback. Our main objective in the research paper was in fact to study the "what-if" situation in which Bitcoin offered the cancellation of the transaction from the user's point of view. Our main interest was the model of the user-agents and quantify the possible "satisfaction" that the user can obtain while also quantifying a possible greater satisfaction with the respect to the current situation. When we wrote the document, we thought about the "implementability" of the cancellation through non-monotonous validity to obtain a more realistic model, but being very cautious in proposing or claiming any kind of mechanism. Of course we never thought that nobody ever proposed it before, that's why when we finished writing the paper, many questions remained unanswered and we decided to send you the document, to get your opinion, which is very useful for improving  the current model.
Although Bitcoin will never implement the mechanism because arguments for non-monotonous validity use case will not emerge, I think this type of study can be useful to conclude on that opportunity or as Brandon suggested to move to other approaches, like Lightning (even though even for Lightning several limitations still there exist !).
    I believe you may be missing the overall points in the "Nail In the
    Coffin" and "Temporary Discussion" sections. In summary:
    1: Any UTXO spending a script with an expiration must be treated
    similarly to Coinbase (I proposed a solution to this, but it's complex
    and may have unforeseen implications).
    2: All existing software assumes that a transaction once valid stays
    valid. Any proposal to change this must ensure that existing wallets and
    users aren't immediately open to being scammed by malicious actors
    sending low fee expiring transactions.
    The more tenable ways to move forward on improving the ecosystem around
    delayed transactions and refunds are: Lightning, improved fee
    estimation, and improved mempool eviction / re-propagation resistance.
    The original reason that I began looking into this is because I noticed
    that during high fee periods, transactions could re-propagate between
    mempools of differing policies resulting in coins being stuck unusable
    for far longer than the expected 1-2 week eviction. I don't know of any
    concrete work going into investigating or improving this.
    HTH,
    --Brandon
    On 2018-09-07 (Fri) at 09:12:40 +0200, Alejandro Ranchal Pedrosa wrote:
    > Hi all,
    >
    > Thank you for the link, and also to Gregory for the remarks. I did not
    > know about this previous proposal. I think the last paragraph of future
    > work is interesting:
    >
    > "It may be interesting to add enhance OP_CHECKSEQUENCEVERIFY
    > <https://github.com/bitcoin/bips/blob/master/bip-0112.mediawiki> to
    > allow outputs that are spendable by Alice until time foo, always
    > spendable by Bob, and spendable by Joe only after time bar, or other
    > such cases"
    >
    > Perhaps it would allow this functionality, while keeping the validity of
    > coins, if the new OP_zzz took an additional argument than suggested,
    > such that the first one is the timelimit for Alice to keep the coin (say
    > in the first 24 hours), and after those 24 hours the ownership goes to
    > the third argument, say Bob.
    >
    > That is, it is not possible to use only specifying the owner in the
    > first 24 hours. Would this be considered harmful?
    >
    > Best,
    >
    > Alejandro.
    >
    > On 9/6/18 10:32 PM, Brandon Smith wrote:
    > > ade a similar proposal about 7 months ago, and documented some of the
    > > discussion points here: